CPM – Corporate Performance Management

The iBonD Series – intelligent Business on Demand

White Paper:Analytical Services in a SOA

Part 1: Vendor Independent White Paper and Reference ArchitectureEnglish Version 4.1 – February 2008

Authors:Dr. Wolfgang Martin Richard NußdorferWolfgang Martin Team, S.A.R.L Martin CSA Consulting GmbHAnnecy München

Sponsored by

Volume 2:

CPM – Corporate Performance Management

CSA Consulting GmbH / S.A.R.L. Martin

CPM White Paper / Dr. W. Martin / R. Nußdorfer 2/15/2008 Page 2

Preface

The present White Paper “CPM – Corporate Performance Management” is the second white paperof the series “iBOND – intelligent Business on Demand”. It describes the business and technicalarchitecture of operational, tactical, and strategic CPM. CPM is defined as a model enabling abusiness to continuously align business goals and processes and keeping them consistent. CPMworks as a closed-loop model for managing the performance of business processes on theoperational, tactical, and strategic level, i.e. planning, monitoring, and controlling. From a businesspoint of view, this is one logical model, but from a technological point of view, rather differenttechnologies from vendors with completely different roots are clashing together: Traditionalbusiness intelligence vendors meet business integration vendors. The convergence happens viathe model of a service-oriented architecture (SOA).

CPM is also called “business performance management - BPM”. These two terms are absolutelyequivalent. We prefer the term CPM, since the abbreviation BPM has multiple meanings, e.g.,business process management and business process modeling. We will always use the term CPMin this white paper, and readers used to the term “business performance management” shouldalways understand this equivalence. We will also use the term BPM, but in this white paper, BPMwill always mean “business process management”.

Traditional Business Intelligence Vendors

These are vendors who used to act in the market of traditional business intelligence tools and datawarehouse and who have evolved into CPM vendors. First, these vendors addressed tactical andstrategic CPM. Today, they are moving to operational CPM (sometimes called BAM – businessactivity monitoring). Furthermore, first BI solutions and products are coming to market that are SOAbased and publish and consume web services.

Traditional Business Integration Vendors

Vendors of integration platforms (see Nußdorfer, Martin, 2007) have started to offer firstoperational CPM solutions using the labels BAM and PPM (process performance management).Their challenge is to put CPM metrics into the classical financial business context. For doing so,service-orientation is the prerequisite.

Best of Breed Products

In addition to the holistic approaches of the vendors of the two camps, there are specialists offeringbest-of-breed tools and technologies. These products are especially interesting when businessintegration platforms do not include operational CPM features, but provide standardized interfacesto an open process warehouse for accessing all logged data of all process instantiations...

Goal of the CPM White Paper

Enterprises developing CPM solutions will have to decide which basic platform to choose for CPMin the context of iBonD and which additional best-of-breed-products will be required. The focus ofthis series of White Papers will be to assist any decisions in the described environment.

Both authors have lots of experience in IT Business in Management functions as analysts and

CSA Consulting GmbH / S.A.R.L. Martin


business-oriented project leaders. They both have many years of practical experience inaddressing and dealing with strategic deliberations and future developments.

The present CPM-White Paper is divided into two parts. First there is this general part 1 describingthe concepts and facilities of CPM as well as its reference architecture. Part 2 describes vendorplatforms and solutions for CPM architectures. To enable readers getting a fast survey of theactual market, the authors have created a separate description for selected vendors. The followingwhite papers are already available:

arcplan, Cubeware, epoq, Informatica, in-factory, Panoratio, SAP, Spotfire

Version 3.0 of this white paper was published in August 2006. Version 4.0 (August 2007) is acompletely reworked and updated version. The authors will be delighted to receive readerfeedback, commentary, criticism - and of course compliments! In Version 4.1 we welcomePanoratio and StatSoft as new sponsors. This gives us the opportunity to update again and toextend chapter 8 in order to take into account the additional mergers and acquisitions that tookplace in the mean time since version 4.0.1 in November 2007. Furthermore, we added a newchapter 4.3 on text analytics, and updated chapter 6 at the end.

Munich, February 2008 Annecy, February 2008

Richard Nußdorfer Dr. Wolfgang Martin

Managing Director CSA Consulting GmbH Wolfgang Martin Team

Wolfgang Martin Team/CSA Consulting


The authors’ biographies:

17 © 2007 S.A.R.L. Martin

Dr. Wolfgang MartinDr. Wolfgang MartinBiographyRecently designated one of the top 10 most influential IT consultants in Europe

(by Info Economist magazine), Wolfgang Martin is a leading authority onCustomer Relationship Management (CRM), Business Process, Rules, andMaster Data Management (BPM/BRM/MDM), Business Intelligence (BI),Corporate Performance Management (CPM), and service orientedarchitectures (SOA). He is a founding partner of iBonD Ltd, VentanaResearch Advisor, and Research Advisor at the Institute for BusinessIntelligence at the Steinbeis University, Berlin.

After 5½ years with META Group, latterly as Senior Vice President InternationalApplication Delivery Strategies, Mr. Martin established the Wolfgang MartinTeam. Here he continues to focus on technological innovations that drivebusiness, examining their impact on organization, enterprise culture,business architecture and business processes.

Mr. Martin is a notable commentator on conference platforms and in TV appearances across Europe.His analytic skills are sought by many of Europe’s leading companies in consulting engagements.A frequent contributor of articles for IT journals and trade papers, he is also an editor of technicalliterature, such as the Strategic Bulletins on BI, CRM and EAI (www.it-research.net), as well as“Data Warehousing – Data Mining – OLAP” (Bonn, 1998), and "Jahresgutachten CRM",(Würzburg, 2002, 2003, 2004, 2005 & 2007).

Prior to META Group, Wolfgang Martin held various management positions with Sybase and SoftwareAG, responsible for business development, marketing and product marketing. Prior to this, hebecame an expert on decision support while with Comshare. His academic work includedComputational Statistics at the Universities of Bonn (Germany) and Paris-Sud (France).

Dr. Martin has a doctoral rer.nat. degree in Applied Mathematics from the University of Bonn(Germany).


BiographyRichard Nussdorfer has worked for more than 30 years in the IT-industry

as a software architect and business analyst.His current focus is on modernization of IT architectures through Business

Integration based on service oriented architectures (SOA) and end-to-end business processes.

Richard’s technical knowledge has been used extensively for integrationprojects, modernizing IT-Architectures, and re-centralizingClient/Server-Architectures to Web-Architectures.

He has published 2 e-books: Information-Technology and the EAI-Book. He regularly contributesarticles to IT journals and is asked to speak at numerous congresses and seminars such asBusiness Integration, DataWarehouses and Business Processes.

Richard Nussdorfer’s professional experience started in 1970 at Siemens AG in softwaredevelopment. He then continued as an expert on databases and project leader for databaseprojects, nationally and internationally , from London to Moscow and from Stockholm toJohannesburg.His professional career continued as manager for Software-Marketing in Munich and BusinessDevelopment Manager in South Africa.From 1990 to 1993 he worked as a consultant for Plenum AG in strategic IT-projects.In 1994 he founded CSA Consulting GmbH where he works today as Managing Director.

Richard Nussdorfer has a degree in computer science from the Technical University in Vienna(Austria).

Richard NussdorferRichard Nussdorfer

http://www.it-research.net),



Copyright

CSA Consulting GmbH/Richard Nußdorfer and S.A.R.L. Martin/Dr. Wolfgang Martin authored thisreport. All data and information was gathered conscientiously and with the greatest attention todetail, utilizing scientific methods. However, no guarantee can be made with regard tocompleteness and accuracy.

CSA Consulting GmbH and S.A.R.L. Martin disclaim all implied warranties including withoutlimitation warranties of merchantability or fitness for a particular purpose. CSA Consulting andS.A.R.L. Martin shall have no liability for any direct, incidental special or consequential damage orlost profits. The information is not intended to be used as the primary basis of investmentdecisions.

CSA Consulting GmbH and S.A.R.L. Martin reserve all rights to the content of this study. Data andinformation remain the property of CSA Consulting GmbH and S.A.R.L. Martin for purposes of dataprivacy. Reproductions, even excerpts, are only permitted with the written consent of CSAConsulting GmbH and S.A.R.L. Martin.

Copyright © 2004 – 2008 CSA Consulting GmbH, Munich/Germany and S.A.R.L. Martin,Annecy/France

Disclaimer

Reference herein to any specific commercial products, process, or service by trade name,trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement,recommendation, or favoring by CSA Consulting GmbH and S.A.R.L. Martin.



Contents

1 Management Summary 7

2 Metamorphosing Business Intelligence 10

2.1 Pitfalls of Business Intelligence 11

2.2 The New Paradigm of Process-Orientation 12

3 CPM – Strategies, Processes, Men and Metrics 18

3.1 Analytics: Process-Oriented Business Intelligence 18

3.2 The Process Ownership Model 23

4 CPM – Methods and Technologies 26

4.1 CPM Business Components 26

4.2 From Business Intelligence to Business Analytics 28

4.3 Text Analytics 30

4.4 Analytics in a SOA 31

4.5 Analytical Services 33

5 Data Integration 38

5.1 Data Integration Platform 38

5.2 Information Services 39

5.3 Meta and Master Data Management 41

5.4 Data Quality 44

6 Latency matters 47

7 CPM and classical BI: fundamental differences 50

8 Players in the CPM/BI Market 51

9 Summary 55

10 The Sponsors 56



1 Management SummaryIn economical down times, budgets become tighter and tighter. Indeed, taking wrong decisionstoday ends in disasters. Identifying potentials for profit, rigorously cutting cost as well as preciselycalculating where to optimally spend the remaining resources are key issues not only for topmanagement. Geopolitical uncertainties make planning much more difficult, but more importantthan ever. New regulations like the Sarbanes-Oxley Act in the US, the International FinancialReporting Standards (IFRS) in the EC, for banking Bale II, and for insurance Solvency II impactfinancial reporting and consolidation. What is the next strategic move to master these challenges?

One answer is Corporate Performance Management (CPM), the topic of this second white paper ofthe iBonD (“intelligent business on demand”) initiative. iBonD explains what makes up winners inthe markets.

Winners do:

• Focus on customers

• Strip away low value activities

• Decentralise decision making

• Drive for compliance

• Industrialize processes

• Collaborate with suppliers, partners, customers

• Focus on agility and are empowered to follow strategic moves on the spot

• and adopt corporate performance management according to the leitmotiv:

You can only manage what you can measure

To summarize: Processes make up the competitiveness of an enterprise. Winning and loosing inthe global market depends on quality and flexibility of business processes. Processes become thenew focus of management (see Nußdorfer, Martin, 2007). Winners in the markets industrialize theirbusiness processes and make them agile. Agility means the power to innovate and tocontinuously adapt its business models and processes to a steadily growing market dynamics. Lifecycles of business processes get shorter and shorter. As a consequence, speed of changes mustincrease faster and faster. Drivers for industrialisation are continuous optimization and higherprofitability. Industrialization means automation and standardization. It speeds up and increasesthroughput as well as it improves quality.

For today’s enterprises, agility and industrialization are key differentiators making up winners orlosers in the world’s global markets.

Business Process Management (BPM) is the answer, a closed-loop model describing the life cycleof business processes, from analysis and design via flow and execution to planning, monitoring



and controlling. The task of CPM within BPM is planning, monitoring, and controlling ofprocesses and their performance.

For CPM and BPM, an appropriate IT support through the right infrastructure is essential. Aservice-oriented architecture (SOA) is required as an infrastructure for closed-loop management ofbusiness processes. BPM and CPM on a SOA enable and empower automated, standardized,reliable, audit-proof, and flexible processes across business functions, departments and evenacross enterprises. This cuts cost and boosts revenues. SOA based processes are independent ofthe underlying IT systems and applications. Hence, business can change processes with thespeed of market dynamics and customer needs. You keep sailing close to the wind. The challengeis continuous adaptation of strategy and processes to market and customer demands; andmoreover...processes must be "intelligent". Through a SOA, analytics can be embedded intoprocesses. Analytics is key for planning, monitoring and controlling processes and theirperformance. The mission is: to identify problems in the right time to take preventive actions.

An example from day-to-day life explains how predictive models work: In a departmentstore, the sales areas are stocked up at the right time, before products are out of stock.This avoids the situation where a customer wants to buy a product and finds himselfstanding in front of empty shelves.

In a process oriented enterprise, CPM and BPM must go together. CPM is the business modelenabling a business to continuously align business goals and processes and keeping themconsistent. The concept is metrics-driven management, the methodology is CPM, and thetechnology is business analytics.

Corporate performance management is fundamentally different from the traditional businessintelligence approach for decision support, executive information, and reporting. Integrated,embedded analytics is the next step to go beyond BI. Traditional BI tools (reporting, adhocquerying, OLAP – online analytical processing, data mining etc.) failed to deliver the rightinformation to the right location in the right time for the right purpose. Traditional businessintelligence tools did not meet management expectations: results to be applied to processes andstrategy for turning information into value. Return on investment (ROI) in the old tools was typicallyrather low, if measurable at all. Traditional business intelligence tools were difficult to master.Information remained a privilege in many enterprises. Only a handful of experts (the power usersor business analysts) were in a position to exploit information via the old tools. Managementdecisions and actions were based on guesses, much less on facts. Embedding analytics intoprocesses through a SOA overcomes these problems.

A SOA enables adaptability and flexibility by separating process logic and flow from business andapplication logic. It is service-oriented, plus it includes a common business vocabulary across allservices. SOA enabled processes can act, not only react. Events can drive process logic and flow.

Example: Product availability should control product publication in a web shop. This preventscustomers to order products that are out of stock and helps business to retain customers andto keep revenues by offering substitute products in the shop.



Corporate performance management is a new approach based on business intelligence foroptimally planning, monitoring and controlling business processes and their performance on thelevel of operations, tactics, and strategies. CPM is based on metrics associated with theprocesses. CPM starts when designing and engineering processes: metrics have to be derivedsimultaneously and in parallel with the operational process design. Goals have to be metricized.Achievement of goals has to be continuously monitored. Actions must be taken for controlling theperformance of processes. CPM is a closed-loop model.

CPM provides clear benefits to an enterprise:

• It is a methodology to link strategy to results.

• It turns data into actionable information.

• It empowers all staff by delivering information not only to power users and business analysts,but to everybody inside and outside the enterprise (“information democracy”).

• It delivers high degree of accuracy and consistency of information.

• It provides transparency to management and enhances the bottom line.

• It delivers the right information to the right information consumer to the right location in due time(This is called “real-time”).



2 Metamorphosing Business IntelligenceWhat is Business Intelligence (BI)? What is its exact definition? Gartner Group used this termalready in 1993, but even today, the term is by far not really known by everybody in business. Just50% of all enterprises use BI today. So, let us start with a definition of BI that is rather widelyaccepted by the market:

Business Intelligence means the capacity to know and to understand as well as the readiness ofcomprehension in order to exercise this knowledge and understanding for mastering and improvingthe business. In somewhat more detail, we define: Business Intelligence is a model consisting of allstrategies, processes, and technologies that create information out of data and derive knowledgeout of information so that business decisions can be put on facts that launch activities forcontrolling business strategies and processes.

The idea of BI principles and concepts is all about to put decisions on facts and to make “betterdecisions”. BI should give answers to questions like

• Do you know which of your suppliers is mission critical to your production? Will their failurebring down your production for hours or even days?

• Do you know what percentage of supplier revenue is due to your spending? Do you get goodterms and conditions from suppliers, using this information?

• Do you know who your most profitable customers are? Are you providing superior services inorder to retain them and are you able to service them, up-sell/cross-sell at appropriate pointswhen interacting with them?

• Do you know in Q1 that you will miss your sales target in Q4, because your actual volume ofleads is insufficient?

• Do you know what revenue you are actually loosing because customers cannot connect to yourcall centre due to peak demand?

• Do you know how much business you miss by not fully exploiting cross-sell opportunities inface-to-face encounters, outlets, and web shops?

Do you know how much money this means for your enterprise? Do you know how to find it, get itand keep it?

We practise BI since 15/20 years. Did we all get the answers we have been expecting to get fromBI? The problem has been since the early days in BI to use its concepts and principles in the dailyoperational business and to get answers that matter. How are we doing?



2.1 Pitfalls of Business Intelligence

Up to now, business intelligence enabled decision support in the context of strategic planning andtactical analysis. The goal of traditional BI was to base decisions on facts. Unfortunately, in manycases this did not deliver the expected added value and enterprise wide acceptance. Reports,indicators, analytical applications and others: where is the real value? Indeed, BI tools failed todeliver. It was always difficult to measure value achieved by business intelligence and the datawarehouse. Reason is: information per se does not create any value. Value is created, wheninformation is applied, used and turned into decisions and actions.

What was wrong with traditional Business Intelligence?

• Business Intelligence was bottom-up and not process-oriented. Line of business people werenot sufficiently involved. Genuine, process-oriented business requirements had not beenaddressed at all. BI lacked business-oriented relevance.

• Business Intelligence was just an information access model for decision support (i.e. BillInmon’s “Information Factory“; Inmon, 1996). This means, information and the analyticalprocesses for information exploitation were mashed together. The results are inflexibility andunnecessary complexity. Any innovation gets discarded from the very beginning. Acceptancedecreases drastically.

• Business Intelligence did support decision making to a certain degree, but the feedbackcomponent for closing the loop was missing. Taking actions based on decisions was not part ofthe model. Indicators that are not in the context of a process bring only limited value. The realvalue of information is only achieved when information is deployed in the context of processes.

Example: As soon as an indicator on the strategic level is in red, the owner of that indicatorhas to make decisions for launching tactical and operational actions. Information isdeployed, decisions are based on facts, and a much higher value is achieved than with thetraditional BI model where feed back is not part of the model.

• Operational aspects of Business Intelligence were left out. Traditional BI was based on a datawarehouse as the single point of truth. This architecture excluded BI from use in operationalenvironments. BI was isolated and limited to tactical and strategic analysis. The potential ofreal-time analysis was completely neglected.

• Business Intelligence was retrospective. Focus was on analysis and diagnostics only. Thepotential of predictive models for identifying of problems and risks in the right time was ignored.

Example: A midrange manufacturing plant analyses quality of production at the end ofeach shift. This guarantees that problems in production are identified as soon as possibleso that actions can be immediately taken to ensure. This ensures that the identifiedproblems will not occur in the subsequent shift. Pro-active BI creates significant value to beexploited.



• Business Intelligence tools did not supply the information consumer and manager sufficiently.Either information was not accessible (or even hidden and retained), or there was aninformation deluge. Again, this lowered acceptance of BI dramatically.

• Business Intelligence was a tools-centric approach based on proprietary technologies. Eachanalytical component played its own role in an isolated environment. Incompatibility andinconsistency were the consequences, and stove-piped information silos were the results. Onthe board level, numbers did not match any more.

Business Intelligence has to be reinvented. The old idea of basing decisions on facts is not badat all. What should and could to be done?

2.2 The New Paradigm of Process-Orientation

Business Intelligence has to be put into the context of processes for achieving business relevance.

What are the relevance and importance of business processes? What can be achieved byprocess-orientation? To get an answer, let us look back: In the 90s, it was common belief thatenterprises could run exclusively on a single instance ERP application. Enterprises becameapplication oriented. Ideally, all business-relevant data was meant to reside in a single databaseand all business functions were meant to have been supported by standard (ERP) functionality.Unfortunately this ideal world was never achieved. What lessons have been learned?

• “One size ERP application fits all” does not work. The majority of enterprises run severalheterogeneous instances of ERP plus legacy and other systems. Enterprises have an averageof 50 mission critical OLTP systems.

• IT performance suffers. The huge number of point-to-point interfaces necessary to linkapplications drives up costs for implementing new applications. The budget for maintainingthese interfaces killed IT innovation. IT became a legacy.

• Process automation is minimal to non-existent. Data has to be manually re-entered fromapplication to application. This makes process quality low and results in mistakes, failures andlost money.

• Process integration is modest to non-existent. Processes end at the boundaries ofapplications making collaboration with suppliers, partners, and customers impossible. As aresult, enterprises are sluggish and unable to react to changes in the market. Costs are drivensky-high.

• Changing your strategy and adapting your business processes to the speed anddynamics of the markets is impossible. Because business processes are hard-coded in theapplications, if you need to change the business process, you need to change the applicationand every other application with which it interacts. In consequence, IT dictates the business,not strategy. This is not practical. Application-oriented enterprises are not agile and willultimately lose to the competition.



• Master data is caught in applications. Each application has its own business vocabulary.Product or order numbers are defined completely differently from one application to the next.Collaboration across networks of suppliers requires master data translation. Each time you adda new supplier, customer, or product you must create new translation tables and/or add thenew item to all the translation tables. This makes changes slow, error prone and costly.

• Information management is impossible. Timely access to business information acrossapplication islands becomes a luxury enterprises can’t afford. The price of not having access tobusiness information is even higher.

How can the traditional enterprise be transformed from an application-centric focus to process-centric model? The answer is Business Process Management (BPM).

BPM is a closed-loop model consisting of three phases (Fig. 1):

Phase 1: Analyzing, planning, modeling, testing, and simulating business processes

Phase 2: Executing business processes by cross-application process flows through a processengine on a SOA (service oriented architecture) infrastructure

Phase 3: Planning, monitoring and controlling of processes and the performance of the ensembleof all business processes

To summarize, BPM means closed-loop management of business processes. It enablessynchronization of execution and exception management with continuous and comprehensiveplanning, monitoring, and controlling. This synchronization keeps business processes optimized inline with real time events and intelligent planning and forecasting.

Business processes are becoming the common communication platform between business and ITpeople. For the first time we can create a genuine dialogue between business and IT. Thebenefits of process-orientation are obvious:

• Processes become the common communication platform between business and IT. Thespecification of business requirements is now based on a common language jointly understoodand spoken by the two parties, business and IT. Technical design of executable processes andback-end services providing application logic becomes straightforward when based on acommon business design of processes.

• Processes become independent from applications. Collaboration makes enterprises shift toend-to-end processes across applications and platforms that are executed by rules-basedprocess-engines running on an integration hub for application and data, the infrastructure forbusiness process management and service management. An important point is that we arenow dealing with cross-functional, cross-departmental, and even cross-enterprise processesthat exploit the application logic of the existing application landscape.




SOA as Infrastructure

The Process-Oriented EnterpriseThe Process-Oriented Enterprise

collaborative business processcollaborative business process

collaborative business processcollaborative business process

Business Process Management

Plan, Monitor& Control

Metrics,BusinessAnalytics

ModelAnalysis,

Design, Test,Simulation

ExecuteRules Based,ApplicationIndependent

Process Engine

CorporatePerformanceManagement

Figure 1: Business Process Management (BPM) is a closed-loop model. Management of businessprocesses becomes the center point of all entrepreneurial actions and activities. Processes are modeled,executed, planned, monitored, and controlled independently of the existing application framework. Theinfrastructure is a SOA (service oriented architecture). Corporate Performance Management (CPM) is asecond closed-loop model for managing the planning, monitoring and controlling of business processes andtheir performance within BPM. This process-orientation is the foundation of an intelligent and agile real-timeenterprise.

• Processes benefit from the advantages of service-orientation. A SOA is business-driven.The granularity of the process model determines the granularity of business services managedin a SOA. Furthermore, the SOA maps technical services from existing back-end applicationsto business services. This is 100% protection of investment in the existing IT architecture. Withservice-orientation we do the next step and build on top of the existing IT investments.

• Processes run across the underlying application data models. In order to automate event-driven processes across functions, departments, and enterprises, commonly-used applicationtouch-points and data across the enterprise must not only be integrated and synchronized, butdata models must be aggregated into a common information model to support collaborationprocesses. This common business vocabulary is the heart of master data management.Uniquely defined and centrally managed ‘meta’ data provides a common platform for allbusiness terms and items across different applications and business constituents. This isessential when defining new products, gaining new customers, or adding suppliers to thebusiness network. One simple update in the master database propagates changes safely andautomatically to all related systems and services.



• Processes consume and publish services. The shift here is from application-orientedthinking to SOA-enabled processes (Fig. 2). For a specific business process, operational,analytical, collaborative and information services are composed by a rules-based processengine. The result is that a business process either becomes a service or a group of services.Certain re-usability can be achieved by avoiding redundant implementation of functions anddata. Redundancy was inherent in the old application-oriented model; service-orientation helpsto overcome this problem.


Business Analytics in a SOA

CAD/CAM

OfficeApplication

ContentManagement

Meta Data Mgt

Intelligence &PerformanceManagement

ProcessPortal

BPM

B2B

DIDI

SOA means• IT Architecture• Enterprise Architecture• Collaboration Architecture

Integration HubESB

ERPCRMSCMPLMDW

legacyetc

Backend Services

Operational D

ata

Market Place, Suppliers, Partners,Dealers, Customers, Social MediaMarket Place, Suppliers, Partners,Dealers, Customers, Social Media

Presentation &Collaboration Services

Figure 2: A SOA describes the design of the infrastructure for BPM. The implementation is based on anintegration hub supporting the life cycle management of processes and managing the back-end servicesincluding information services (DI = data integration) and meta/master data services. It also provides the B2Binterface. Other business domains like content and knowledge management, office and CAD/CAM can alsobe incorporated via the integration hub. Corporate Performance Management acts as the brains of theprocess-oriented enterprise. It provides the “intelligence” for optimal monitoring and controlling all businessprocesses and their performance. Analytics is embedded into the processes for anticipating problems andrisks. The Process Portal acts as the human interface. It supports human interactions through collaborationand presentation services. A Process Portal supports multi-channel communication via the web, PDAs,voice, etc. (ESB – enterprise service bus, ERP – enterprise resource planning, CRM – customer relationshipmanagement, SCM – supply chain management, PLM – product life cycle management, DW – datawarehouse, B2B – business to business)

• Processes drive the transformation to intelligent real-time enterprises. Businessintelligence is gleaned from metrics associated with each business process. Business metricsare defined by goals and objectives to manage a process and its performance in a measurableand proactive way with information, key performance indicators (KPI), rules, and predictivemodels.



Example: Let us assume, term of delivery is a goal of the shipment process. Then we firsthave to make the goal measurable. As a metric, we could define that 90% of all shipmentsshould be within 2 days. This is a strategic metric. An operational business metric could bea predefined threshold for stock in a dealer warehouse. If stock falls below the threshold, anorder is automatically executed. The outcome of this metric on stock level launches anaction. It is a pro-active metric to avoid the problem of sold out.

Metrics can be anticipative as this example shows. In CPM, we go beyond traditional diagnostics.Based on anticipative metrics, processes get the power to act proactively and to become “selfhealing”: Problems and risks are identified in the right time, and decisions and actions are taken toprevent damages. In other words:

We have reinvented Business Intelligence. We have put Business Intelligence into thecontext of business processes. Business Intelligence becomes Business Analytics.

Business Analytics is about planning, monitoring, and controlling of business processes and theirperformance. This model is called Corporate Performance Management (CPM).

Definition: In a process-oriented enterprise, CPM is the model enabling a business tocontinuously align business goals and processes and keeping them consistent. CPM meansplanning, monitoring, and controlling of processes.

Finally, we have to define the infrastructure for BPM and CPM so that we can embed analytics intoprocesses. As we have already seen (Fig. 1), this is done through a SOA (Fig. 2). From the ITpoint of view, agility and industrialization are two contradictory requirements, but if theinfrastructure for managing processes is a service oriented architecture (SOA), then the twoprinciples are brought together. The reason is the nature of a SOA. It is a special architecture forproviding “Software for Change”. This is due to the following principles of a SOA:

A SOA is the infrastructure for BPM and CPM that separates process and business logic.

• SOA is a design model for a special enterprise architecture and a special enterprise softwarearchitecture

• SOA is independent from technology

• SOA is an evolution of component architectures (principles of “LEGO” programming)

• SOA services are business driven. The granularity of the process model determines thegranularity of business services.

An architecture following these principles is called service-oriented if the following three principalshold:

Service-Orientation• Principle 1 – Consistent Result Responsibility. The service provider takes responsibility for

the execution and result of the service. The service consumer takes responsibility forcontrolling service execution.



• Principle 2 – Unambiguous Service Level. The execution of each service is clearly agreed toin terms of time, costs and quality. Input and output of services are clearly defined and knownto both parties by the Service Level Agreement (SLA).

• Principle 3 – Proactive Event Sharing. The service consumer is informed about every agreedchange of status for his work order. The service provider is required to immediately inform theservice consumer of any unforeseen events.

Such a service orientation provides a flexible framework for standardizing and automating businessprocesses, for bundling regional and global competencies into service offerings, for load balancingof peaks “on demand”, and for provisioning services by third parties via a “software as a service”(SaaS) model. Services provide business and decision logic that traditionally was included inapplications. Processes

As a prerequisite for applying these three rules, we need a business vocabulary so that all SOAbased processes use the same notation and specifications. A repository is necessary for uniquelydefining all meta and master data. The repository for meta and master data plays a similar role asthe integration hub within a SOA. So, the architecture of the repository should be hub and spoke sothat all meta and master data can be synchronized and versionized across all back-end systemsand services. This is the role of master data management (MDM). In chapt. 5, we will discussMDM in more detail.

Note: ROI is not provided through a SOA, but through the implemented processes.



3 CPM – Strategies, Processes, Men and Metrics

3.1 Analytics: Process-Oriented Business Intelligence

As we have seen in the previous chapter, CPM puts business intelligence into the context ofprocesses. The obvious consequence is CPM also puts BI into the context of strategy and men.Today, processes are cross-functional, cross-departmental, and cross-enterprise. They link thesuppliers of the suppliers with the customers of the customers. Let us recall the definition of abusiness process.

A business process is…a set of activities and tasks carried out by resources

(services rendered by people and machines)using different kinds of information

(structured & unstructured)by means of diverse interactions

(predictable & unpredictable)governed by management policies and principles

(business rules & decision criteria)with the goal of delivering agreed upon final results

(strategies & goals)

The benefits and advantages of integrated end-to-end processes are obvious:

• Faster and more reliable processes cut costs. Automation improves speed and quality ofprocesses. The result is higher throughput with fewer resources.

• Integrating processes shortens time-to-market. The ability to respond quickly to newopportunities, customer needs, market dynamics and problems simply translates into increasedrevenue and profitability.

• Safe and reliable processes minimise risk. High process quality means less costly aftershocksto the bottom line. In addition to savings realised by reductions in post-sales service,enterprises can benefit from high customer satisfaction and, ultimately, market share. Theability to anticipate problems, customer needs, and market dynamics makes the intelligent real-time enterprise a reality.

• Through flexible process management - independent of applications - you maximize businessflexibility and agility. By removing the constraints in hard-coded processes intrinsic ERP- andother standard application packages your processes will move in line with market dynamics.

• Process-orientation creates transparency and traceability. There is no alternative to compliancewith the regulations of public authorities and the requirements imposed by auditors.

This is why “Business Process Management (BPM)” is one of the most important challenges fortoday’s enterprises. BPM and CPM are the process-oriented latest version of managing an



enterprise: planning, execution, and performance management have always been the three basiccategories of all management (“make a plan, execute it, and manage to keep the actual in line withthe plan”).

CPM within the BPM model is all about managing the performance of all processes thatextend across all functions within a business, and beyond to all other relationships in business tobusiness and business to consumer. Metrics-oriented management is the top down principal ofCPM for optimal enterprise management by a closed-loop approach (Fig. 3). Business strategydetermines which business processes are to be executed and managed by the enterprise.Business metrics are associated with each business process. Business metrics are defined bygoals and objectives to manage a process in a measurable way with information, performanceindicators, rules, and predictive models.


CPM: Strategy, Goals, Processes, MetricsCPM: Strategy, Goals, Processes, Metrics

Cycle SpeedCycle Speed

EndResultBusiness ProcessBusiness Process

ActAct DecideDecide MeasureMeasure

Strategy

EventsGoal

Figure 3: Metrics-Oriented Management is a top down model for information-based business management.Measurable goals and objectives are derived from the strategy. Based on strategy, goals and objectives,business processes and business metrics for efficient process control and continuous optimization aremodeled in parallel. Technical implementation of processes and metrics follows the principles of a SOA(service oriented architecture) by operational and analytical services. Based on monitoring, decisions aretaken either manually by man or automatically by decision engines. Decisions lead to actions for controllingthe process and its performance (tactical and operational BPM) as well as updates strategy, goals andobjectives (strategic BPM). The loop closes. Synchronizing monitoring, decision and action taking with thespeed of the business process and business dynamics is key – indeed, this is a foundation of the real timeenterprise.



Embedding analytics in processes requires a new approach to process modeling as well as a newapproach to business intelligence. Modeling process logic and flow only as in the past isinsufficient. We now have to model simultaneously metrics and responsibilities. We have to linkstrategy and goals to processes, metrics, and people and to build the closed-loop. This is all aboutgovernance. Governance means an organization and control of activities and resources in theenterprise that is oriented to responsibility and to durable and long term value creation.

Example: Monitoring and controlling of sales processes. Sales methodologies describe andstructure the sales activities across the sales cycle. The sales cycle is typically defined asthe time period between the identification of a lead and the payment of the bill according toa signed contract. The methodology describes the various levels of qualification of a leadand the actions to be taken to move a lead from one level to the next. These levelscorrespond to the different states of the sales process, where the desired end result andfinal state is the payment of the bill. The number of levels of qualification depends on theselected sales methodology (and does not matter in this example). Now let us apply CPMto this sample process. The metrics for monitoring and controlling of our sales process arethe number of qualified leads per level, the estimated/achieved value of a deal, and thetransition rate and transition time from each level to the subsequent level. Based on thesemetrics, we can now act proactively. Assume that the objective of our sales process is arevenue of x € in six months. We then can estimate the number of leads per level that isnecessary to achieve this goal by evaluating the defined metrics and compare the resultwith the actual sales data. If the result shows that we will not achieve our goal, we still havetime to preventively counteract by taking actions for controlling the process, e.g., anadditional lead generation for filling up the sales funnel.

Metrics-oriented management is based on information management. Information has to beavailable in “right-time” (often called “real-time”, see Nußdorfer, Martin, 2003) for triggering manualor automated decisions for process control. This corresponds to the “information supply chain”paradigm: supply the right information in the right time to the right location to the right informationconsumer to trigger the right decision. So “real-time” means synchronization of information supplywith information demand (Note: “real-time” is a relative term and not necessarily related to clock-time).

Business metrics represent management policies within metrics-oriented management. The ideabehind is obvious:

You can only manage what you can measure.

So, flexibility of changing and updating any metrics is one of the top requirements of the model.Furthermore, business metrics must be consistent. Metrics specified to control the execution of aparticular group of processes should not contradict other metrics. Indeed, metrics are cross-functional and cross-process: The performance of a business process may influence and interferewith the performance of other processes.

For example, delivery time, a supply chain related metric, may influence customersatisfaction, a customer relationship management metric.



These issues are addressed by business scorecards. A business scorecard aligns allmanagement policies presented by all metrics across the enterprise and presents the aggregatedtop management policy of the enterprise as well as all details for all employees. Examples ofparticular business scorecards are Norton/Kaplan’s balanced score card or the six sigma model.The balanced scorecard, for instance, is a collection of metrics that is not only based on financialparameters, but uses also customer, employees and shareholders loyalties to provide a look to thecorporate performance beyond the quarterly results. It presents indeed one particular style ofmanagement policies. Despite the wide variety of these metrics, the final goal remains the same:transform data into information and knowledge and maximize its value for the business by closingthe loop: We now base planning, monitoring and controlling of processes on information, facts, andknowledge.

CPM is applied to all business domains like customer relationship management, supply chainmanagement, human relations etc.

Example: Financial Performance Management like any other analytical solution is aclosed loop process depicting the information management of financial information. Theprocess stretches from planning, budgeting, and forecasting to performance measurementand auditing via financial metrics including the statutory legal financial reporting andconsolidation requirements. Financial performance management includes profitabilityanalysis and planning as well as simulations and what if analysis. Decisions are then madebased on the financial metrics and analysis and fed back into the planning, budgeting andforecasting activities: The loop is closed.

As Fig. 3 already implies, CPM takes place on three levels, the operational, tactical, and strategiclevel (Fig. 4). In the past, business intelligence focused on enabling decision support in the contextof strategic planning and tactical analysis. This was done by metrics designed for long termoutlooks. The basic concepts were to measure and to monitor the achievements of strategic goals,for example customer satisfaction, customer value, term of delivery, supplier value, staff fluctuationetc. Long term here relates to the dynamics of the process. Question is how fast can actionsinfluence the process and significantly change the indicators. This is why there is a tactical level.Achievement of tactical goals can be considered as mile stones towards the strategic goals.Actions targeting the achievement of tactical goals typically address a time frame between somefew days to several months. Today, process orientation operationalizes business intelligence.Operational processes are to be monitored and controlled in right time (“real time”) via intelligence.Operational CPM is also called “Process Performance Management (PPM)”, and it includes“Business Activity Monitoring (BAM)”.

These ideas stem from control theory. As room temperature is monitored controlled by a closedloop feedback model, business processes shall be monitored and controlled on the operationallevel, i.e. real-time. Figure 4 depicts the three levels of CPM. The real-time principles of theinformation supply chain enable monitoring and controlling even of operational systems. Aninformation supply chain is defined by the principle of the availability of the right information in theright time at the right location for the right purpose. Information is treated as the duty of theinformation provider. In the data warehouse model, information was treated as the duty of the



information consumer. In CPM, the provider of information is responsible to propagate information.The implementation is done through a publish and subscribe communication method.

Example. In a web shop, product availability is a valuable metric when controlling the orderprocess. Product availability is an operational metric. It measures stock via sales andsupply transactions. Hence, product availability is synchronized with transactions. Whenproduct availability gets below a certain pre-defined threshold, an alert can be launched.Such an alert could trigger an additional shipment. If shipment is not an option, then theproduct could be blocked in the product catalogue so that customers cannot place anyorders for this product. This is a pro-active action that avoids canceling customer orders. Inthe end, the frustration of customers due to the unavailability of a product is minimized.Furthermore, the blocked product could be tagged by a note stating when the product willbe available again.


CPM – Temporal LayersCPM – Temporal Layers

StrategicCPM – long term

Tactical CPMmid term –

days, weeks, months

Operational CPM (BAM)short term –

same day at least

StrategicPlanning

TacticalAnalysis

OperationalActions

Top downMethodologydriven

Bottom upProjectdriven

Figure 4: Corporate Performance Management (CPM) is the process of managing the performance ofbusiness processes by applying metrics, deciding on the outcome of the metrics, and launching actions forcontrolling the performance and/or the process, a closed loop model. One key issue for all CPM approachesis to put the metrics into a monetary context. This requires process-oriented accounting principles like activitybased management/costing.

CPM spans from operational to strategic CPM, but is addressed by two separate camps of vendors rushingto exploit the new opportunities of a strongly growing analytics market. These are the Business Integrationvendors providing BAM (business activity monitoring) solutions, and the Business Intelligence vendors withtheir traditional focus on tactical and strategic solutions. This is confusing business and IT people looking forreal solutions to solve their more and more complex analytical needs.



This example shows how to monitor and control business processes on the operational level byinformation. Processes are automated; manual interactions of product managers are minimized. Bythe way, what is the meaning of “real-time” in this example? Typically, product availability ismeasured twice a day. This is an empirical experience balancing cost of measuring with cost ofrisk ignoring product availability for controlling the process.

Operational CPM has been addressed first by vendors coming from process engineering andbusiness integration by adding reporting and graphical features for visualizing operationalperformance indicators. Via activity based management and costing these metrics can be also putinto a monetary context. This means technically to have access to financial data in a datawarehouse.

Tactical and strategic CPM was first addressed by the vendors of traditional business intelligenceby moving from the data warehouse model and business intelligence tools to analyticalapplications and closed loop processing. Today, the two independent approaches to one and thesame problem are confusing the market, but are converging since 2005.

3.2 The Process Ownership Model

At this point, we have to address the question of how to use information. There are two aspects tobe considered. On the one hand, we have to find a solution for the information supply chainparadigm, i.e. who needs when, where and why what information? On the other hand, we have tounderstand the skills and training necessary for exploiting information. Let us start with the aspecthow easy to use CPM tools should be.

Traditional BI tools lacked the ease of use features. For users, it was mandatory to have a deepknowledge on how to use report generators, adhoc querying, OLAP tools, spreadsheets, statisticaland data mining tools, etc. This know how was typically acquired by training and education.Business analysts and power users evolved as a new class of people that was empowered by allthese types of tools. Business departments became dependent on this new type of informationempowered employees. So, information became a kind of luxury product that was not available toeverybody. This is now changing in the CPM model. Analytics for performance management andanalytics for enriching operational processes requires that everybody participating in a businessprocess must be in a position to consume information without training and education. This isenabled by the CPM methods and technologies as we will see in the following chapter 4.

Nevertheless, we will continue to engage business analysts and power users, but their role ischanging. Due to better ease of use of CPM tools, business analysts will be less engaged inproviding standard information upon request. So, they can spend more time for interactiveanalytics (data exploration) and create more value for the enterprise. Plus, a new task is attachedto them: management of the CPM methods and technologies. Identifying and communicating bestpractices of analytical scenarios will be their charter. This requires a close cooperation andcollaboration with the information consumers. If an information consumer will be confronted with anew, not yet encountered problem in monitoring and controlling his/her business processes, a new



analytical scenario has to be jointly developed with a business analyst. Once solved, the newscenario can be reused within the CPM framework. The CPM organization learns and gets betterthe longer they apply the CPM model.

Even if CPM provides an intuitive working environment that needs much less training andeducation, there is still the problem of data deluge to be solved. Who needs what information,where, when and why? The solution comes with the information supply chain model linkingprocesses, metrics, people and organization: Process-orientation comes with a processownership model. This is part of the governance of the BPM, CPM, and SOA model.

The process ownership model describes who of the constituents (employees, partners, suppliers,customers etc.) participates in and is responsible for what processes or activities. This enriches theprocess model by the roles and organizational units of all people that are involved in executing andmanaging processes. In metrics-driven management, this process ownership models also includesthe metrics that are necessary to monitor and control the process and its performance. This can beunderstood as information sharing and filtering. The constituents share data, information andknowledge in the context of their process-oriented communication and collaboration. All other datais filtered out. Consequently, the result is a top-down security model as a by product of the processownership model. Information sharing and filtering is done via information profiles describing thecontext of collaboration based on the process ownership model. Some call this “informationdemocracy” (Fig. 5): The process ownership model includes the information profile describing andfiltering exactly the information that is needed by all constituents based on the context ofcollaboration.


CPM – GovernanceCPM – GovernanceO

rganisationO

rganisation

DrivingDriving thetheEnterpriseEnterprise

Process

Process

Ownership

Ownership

Information

Information

Profiles

Profiles

SensorsSensors

PeoplePeoplePeople

MetricsMetricsMetricsProcessesProcessesProcesses Tech

nolo

gyTe

chno

logy

EnterpriseEnterprise StrategyStrategy

Goals &Goals & ObjectivesObjectives

CultureCulture

CPMCPMCPM

CPMCPMCPM

Figure 5: Information democracy comes with the information supply chain paradigm. Everybody has accessto all information necessary to execute and manage the processes and their performance specified by theprocess ownership model – not less and not more.



Visualization of information according to information profiles should be done by portlets embeddedinto portals (see also Fig. 2). Portals have evolved from intranet and extranet solutions to thecentral point of control for collaboration providing the P2S (person to system) interface. A portal isdefined as a system that enables sharing and filtering of data / information, functions / functionality,content / knowledge, and processes. This sharing and filtering is related to the functional role of acollaborative team within the process ownership model. A collaborative team is a group of peoplerepresenting the various constituents that work together according to the collaborative goals andobjectives of the team. In this way, portals support cross-functional, cross-departmental, and cross-enterprise virtual teams. As a special case, a team could also be an individual portal user. Tosummarize, portals enable information democracy.

A process portal (see fig. 2) can be understood as an abstraction layer linking and aggregatingcontents and services as well as reducing the complexity of their access. In this sense, the team-context defines the collaboration bandwidth, i.e. which data / information, functions / functionality,contents / knowledge, and processes are exposed to the collaborative team together with theappropriate collaborative tools. Each portal user gets a personalized environment that can befurther individualized. Indeed, such a person portal can be understood as an integrationtechnology. But the ultimate integration is done via a human interaction, i.e. within the team-context; a user can execute a message transfer between contents and services within his context.

Furthermore, process portals also provide synchronous and asynchronous collaborative tools, e.g.,e-mail, co-browsing, chat, blogs, instant messaging, web-conferencing etc. We have described therole of portals and their relationship with BPM and SOA in Martin and Nußdorfer (2006).



4 CPM – Methods and TechnologiesAs we have already seen, CPM is fundamentally different from traditional BI. Focus of BI was tools,e.g., OLAP, spreadsheets, reporting, adhoc querying, statistical and data mining tools, etc. CPMcomes with new methods and technologies. Goal is to empower everybody collaborating in thecontext of a business process by analytics without becoming a specialist in analytics. This principleis not only applied to employees, but also for suppliers, partners, dealers, and even customers.Analytics must become consumable by everybody.

4.1 CPM Business Components

Metrics and Key Performance Metrics – Metrics are used to manage the performance of aprocess and / or to control a process. They are derived top-down from metricized goals out ofstrategy and process analysis. Metrics work like sensors along the reach of a process flow. Thefinal goal is the proactive identification of risks and problems. Early warnings become possible sothat preventive actions can be taken to bring a process instantiation back on track. (See page 20;example of monitoring and controlling sales processes).

Metrics consist of indicators and scales. Scales define how to interpret instantiations of indicatorsand what decisions to take. A key performance metric (KPM) is a compound, cumulated metric.Term of delivery is an example for a KPM. It is cumulated of detailed metrics like time of deliveryacross all customers within a certain time period. Typically, an employee will have a lot of detailedmetrics, but just some selected KPMs. KPMs should be related to the personal goals and matchthe model of management by objectives. In the end, KPMs could have an impact to certaincomponents of the salary.

In the example about term of delivery as a KPM, a decision maker is responsible for interpretationof the KPM, making decisions, and taking actions. In case of such a human interaction, scales aretypically visualized by traffic lights and / or speedometers. Green, yellow, and red lights ease andspeed up the interpretation of instantiations of KPMs and metrics. In the web shop example aboutmanaging the order process by product availability (p. 22), the interpretation is automated by adecision engine – visualization is not necessary.

Business Scorecard – This is a consistent and comprehensive group of metrics for monitoringand controlling a group of processes or even the total enterprise according to a managementpolicy. Consistency of metrics is very important, because metrics should not be contradictory andcause conflicts between collaborative teams working in different contexts. The term businessscorecard was first developed for strategic CPM, but is now used for all levels of CPM. Knownmodels of business scorecards are the already mentioned balanced scorecard of Kaplan andNorton (www.bscol.com), Baldridge’s scorecard model (www.quality.nist.gov), and the Six Sigmamodel (www.isixsigma.com). It should be noted that he majority of enterprises does not exactly

http://www.bscol.com),

http://www.quality.nist.gov),

http://www.isixsigma.com).



apply one of these models, but uses its own customized scorecard model that is a derivative of oneof these models.


Figure 6: Example of a Strategy Map of a Balanced Scorecard Model built with Actuate/performancesoft. Inthe Strategy Map illustrated here, Input and Process metrics are shown in a cause-and-effect relationship totheir respective Outcome metrics. This pictorial representation of the strategy allows the organisation toevaluate its effectiveness by tracking key measures relating to each corporate objective.

Strategy Maps – Strategy Maps (Fig. 6) are a visual presentation of a strategy based on thecause-and-effect relationship of input and process metrics to their respective output metrics. Thewell-known and typically used indicators of traditional BI were too much biased by financial dataand did not sufficiently consider investments in people, IT, customer relationships as well as insupplier and partner networks. This is why the standard planning and reporting systems like profitand loss and cash flow based on the traditional BI indicators are not applicable for monitoring andcontrolling of resources beyond finance. In genuine CPM, we overcome this problem by theconcepts of process-orientation: We use metrics as sensors and cause-effect relationshipsbetween the various goals and objectives within a strategy. Strategy determines the goals andobjectives of value creation by processes. This is depicted by strategy maps, and the businessscorecards provide the translation into decisions and actions for monitoring and controllingprocesses in a proactive way. Strategy Maps as well as Business Scorecards are not static. Marketdynamics and customer needs drive and change strategy, so strategy maps as well as businessscorecards.



Business Rules – They represent cross process decision logic in the context of businessexpertise and management policies (refer to the definition of a business process on p. 18).Modeling of rules is either top-down by an expert system type approach or bottom-up bygenerating predictive models (e.g., a customer behavior model by a data mining process).Ultimately, rules can be modeled by a combined top-down, bottom-up approach aligning predictivemodels with expert rules. Business rules must be managed centrally and independently ofbusiness processes. The reason is the n : m relationship between rules and processes: a rule canbelong to several processes, and a process can have several rules. When business rules are hardcoded into the processes, then a chaos for maintenance of rules is inevitable after some shorttime, since the consistency of rules will be in danger.

Alerting – Event-orientation requires alerting services. When an event / alert occurs, theinformation describing the event / alert is automatically propagated to all recipients that havesubscribed to receive this information. This is set up in the publish and subscribe communicationmethod using message / queuing infrastructure. The principle of this communication method isdefined by the information supply chain model. All information that is necessary to process theevent / alert should be available to all recipients in right time for making the right decision andtaking the right action. Again, right time means to synchronize the speed of the process with thedelivery of information via the propagation. If speed is high, and the delta between event / alert anddecision / action becomes small, then a human interaction may be to slow: The decision / actiontaking must be automated. Examples for automated decision / action taking can be found onvarious web sites where recommendation engines are working. Rules engines are state-of-the-arttechnology for automated decision taking (see chapt. 6).

Broadcasting – These are services for delivering personalized messages to millions of recipientsvia e-mail, fax, pager, mobile phone, PDA etc. RSS („real simple syndication“) feeds are becomingthe leading technology for data syndication pushed by the more and more widespread usage ofWeb 2.0 concepts. Using exception conditions and recurring schedules as triggers, events can beautomatically created and propagated to processes and people within the enterprise or to anyexternal community. Content can be personalized to the individual subscriber, preventinginformation overload and ensuring that security requirements are strictly enforced.

4.2 From Business Intelligence to Business Analytics

Process-orientation drives the evolution from BI to CPM. The CPM business components requirenew BI tools and services, a new architecture for positioning the tools into the context of CPM (Fig.7) as well as a new fresh thinking in terms of processes and services. We are making the next stepfrom Business Intelligence to Business Analytics. Here come the fundamental differences totraditional BI:

Analytics is process-driven, not data-driven. It links business strategy to processes and peopleaccording to their role in collaborative teams: the use and value of information now goes beyondthe power users and business analysts that in the past were the only people benefiting from



information provided by BI tools. Analytics now empowers all participants of the enterprise valuenetwork, suppliers, partners, dealers as well as customers... It is targeted at the business ratherthan IT.

Analytics can be predictive. It is aimed at responding to unforeseen events and revealing newinsights and unexpected discoveries. It is not limited to the analysis of historical data pre-programmed into a warehouse or a cube.

Embedded Analytics – from strategy to operations. A SOA makes it happen. Embeddinganalytics into operational processes enables synchronising information delivery with process speedand interacting with information at the speed of business so that decisions and actions can betaken in right-time. Through embedded analytics, processes become intelligent and event-driven.


Metricized GoalsProcesses & Metrics

CPM – Reference ArchitectureCPM – Reference Architecture

EmbeddedAnalytics

Strategy

Interactive Analytics

Met

a D

ata

Data Integration

ad hoc WorkflowCollaboration

Adaptive dynamic

AnalyticalServices

Figure 7: Reference architecture for CPM. It enables the comparison of products and offerings of the variousvendors for planning / developing, executing and managing CPM. Key is the coupling of modeling ofprocesses and metrics as well as the top-down implementation of metrics by analytical services and bottom-up by interactive analytics (data exploration). Data integration is the foundation for CPM. It provides paralleland simultaneous access of operational and analytical data via services within the framework of the SOA.

Analytics needs data integration – Traditional BI tools worked exclusively on the datawarehouse. The data warehouse provided the “single point of truth”, i.e. reliable and high qualityinformation. This prohibited the application of BI to operational environments. BI just operated inthe domains of strategic and tactical analysis. The potential value creation by real-time analyticswas discarded. Analytics has parallel and simultaneous access to operational and analytical dataand information. A data integration platform (“enterprise service data bus”) now becomes the single



point of truth. In a SOA, it links CPM to the integration hub and the data warehouse. The dataintegration platform provides information services (Fig. 8) – may be as web services – that can becomposed out of any operational and data warehouse data. The data warehouse becomes abackend service (Fig. 2) providing in particular historical data.

Interactive Analytics – analytical processes and collaboration. Interactive analytics (alsocalled “Data Exploration”) is an adhoc, dynamical, easy to handle, analytical, collaborative process.The goal is to provide new analytics, e.g., profiles, rules, scores, and segmentation for a betterinsight into markets, customers, risks etc. In this sense, it is a bottom-up development environmentfor metrics and predictive models. Good example for interactive analytics is the development ofpredictive models by data mining. The final predictive model is then implemented in a rules enginecontrolling an operational process.

Example. Let us consider the process of credit approval in banking. Standard rules forchecking a customer situation for solvency and credit approval can be rather easilymodeled by a financial consultant. This top-down model can be complimented by a bottom-up model describing the risk of credit failure. This can be identified by data mining customerdata and providing a risk based customer segmentation. A combination of the expert rulesand the generated predictive model provides the final rules. The process of credit approvalcan now be automated, its workflow is controlled by a rules engine, and customers can nowrun credit approval as a self service on a web site, for instance.

Other examples can be found in the context of cross/up-selling and customer attrition. It isimportant to note that special knowledge how this intelligence works is not necessary whenworking in the context of intelligent processes. Analytics embeds intelligence into the process, andworks as a black box. So, Analytics including even sophisticated approaches like data mining, textmining and web mining is made consumable for everybody, not only for some thousands ofspecialists, but for millions and more information consumers.

Business Analytics enables intelligent processes. Operational processes can now be enriched byembedded analytics and can be monitored and controlled “in real-time” (BAM). Service-orientationeases the embedding of analytics. Traditional BI tools turn into analytical services. In a SOA, CPMis implemented through analytical services.

4.3 Text Analytics

Text analytics is a new type of analytics1. It combines linguistic methods with search engines, textmining, data mining, and machine learning algorithms. Text analytics is both, a technology as wellas a process for knowledge discovery and extraction in unstructured data. The first goal of textanalytics is to identify selectively entities (like names, data, locations, conditions) and theirattributes as well as relationships, concepts, and sentiments between entities. The second goal is

1 See also http://www.intelligententerprise.com/blog/archives/2007/02/defining_text_a.html

http://www.intelligententerprise.com/blog/archives/2007/02/defining_text_a.html



to create and to visualize classifications based on the identified structures. As an example, anoutcome of text analytics could be the identification of opinion leaders in social networks.

Text Analytics is the extension of business analytics and data mining supporting analytics inContent Management.

Text analytics is pushed by the increasing adoption of Web 2.0 principles. Given the additionaldata exhibited in social networks, Web 2.0 enables enterprises to address customer segments withchirurgical precision, but it also bears uncontrollable risks: In umpteen thousands of blogs andforums, people talk and chat in all details about products and enterprises – real lies included.Expert forums can demystify quickly and lastingly slogans and claims. Comparisons of pricescreate transparency at cyberspeed. Competitive Intelligence is no more restricted to strategiccompetitor observation, but moves to operational observation of brand new competitor productseven before they come to market. For instance, Nokia was well informed in advance about weakpoints of Apple iphones by expert discussions on certain forums. This all is due to applied textanalytics.

Text analytics is also successful when it is about the identification and classification of criticalcustomers. Critical customers could be very helpful in removing product flaws, but could also benotorious grouches and wiseacres.

Example: BMW actively uses blogs. Experience with bloggers has shown that customerscommunicate sometimes more positively about BMW products than BMW’s own sloganswould dare to. (Attention: Sony has once tried to influence bloggers and blogs. When thiswas brought to light, damage to Sony’s image was serious.) BMW created “M Power World”,a social network about sportive driving for the special customer segment of buyers of Mmodels. Here, customers are invited to exchange ideas with BMW developers and designers.Customer becomes product developer – this is the fundamental Web 2.0 principle.

BMW applies a Web 2.0 forward strategy: Web 2.0 principles become part of their CRM strategy.An alternative would be a passive strategy by automated observing of selected blogs and forumsby text analytics for identifying critical situations and mood changes as quickly as possible. This isvery well doable by text analytics, but it turns out that it is extremely difficult to launch the rightactions in case of. You may legally enforce the deletion of blog entries, but in reality, they will popup elsewhere. In the world of Web 2.0, the principle of “semper aliquid haeret” is inexorable. Here,we are entering virgin soil and have to learn of lot.

4.4 Analytics in a SOA

We have already introduced the concept of a SOA. Cross departmental and cross enterpriseprocesses can be implemented as composite applications – see also Nußdorfer and Martin (2007)– orchestrating business services according to the process logic. Business services present andpublish the business logic from existing back-end systems (see also Fig. 2), or have to bedeveloped and/or acquired, if the necessary business logic has not yet been implemented.



There are five categories of services providing business logic (Fig. 8):


Analytics in a SOAAnalytics in a SOA

OperationalData

DataWarehouse

ExternalData

UnstructuredData

Collaborative Business ProcessCollaborative Business Process

Data IntegrationPlattform

AnalyticalServices

CollaborativeServices

OperationalServices

InformationServices

Third PartyServices (SaaS)

AccessServices

ApplicationServices

Dev

elop

men

tSe

rvic

es

IT M

anag

emen

tSe

rvic

es

Infrastructure Services

Enterprise Service & Service Data BusServicesRepository

RulesServices

Figure 8: Business processes orchestrate services within a SOA. The main idea of service orientation is tosplit process and business logic. There are five categories of services providing business logic, information,analytical, rules, operational, and collaborative services. These categories of services can be considered as“business services”. They are composed out of “Technical Services” provided by 3rd parties (SaaS – softwareas a service), backend applications, and the various types of data sources. Furthermore, we needdevelopment services for both, process logic and business logic, and IT Management Services foradministration, execution and security of services. The Enterprise Service Bus together with the EnterpriseService Data Bus is a kind of intelligent middleware enabling service and data brokerage. It also includes theservice directory listing and publishing all available services.

• Operational Services. They provide transactional business logic like creating new customer,new account, placing an order etc.

• Collaborative Services. They provide services supporting human interactions and person toperson communication like setting up a meeting, search services, communication services likeembedded e-mail, chats, SMS, voice etc.

• Rule-Services. Rules define the decision logic. A process typically uses several rules,whereas a rule can be used in several processes. This is why we have to strictly separateprocess and decision logic. In a SOA, rules are considered as rules services that areorchestrated by the process engine as any other service. A rule service can be alsounderstood as an encapsulation of complex rules. Indeed, a rule service could use anotherrule service as a sub-rule.



• Analytical Services. They provide analytical business logic like a threshold for productavailability, a predictive model for customer behavior or customer risk, a forecasting service forsales, etc.

• Information and Data Services. They provide composite information based on structured andunstructured, operational and analytical data sources like customer address, customer value,term of delivery etc. Information and data services also include meta data and master dataservices.

In this white paper, we now focus on analytical services (chapt. 4.4) and information and dataservices (chapt. 5). Before doing so, let us put another note to rules services. They can also beused to automate human decision making by using rules engine as a decision engine. A decisionengine should have scheduling features for follow up of events by intervening actions. Forinstance, if a customer has visited a web site, given a positive response, but did not come backwithin a certain amount of time, then the decision engine should be able to detect this “non-event”,and send a trigger, for example to a call center agent for follow up. Decision engines enableintelligent interactions with all business constituents. For example, they can enable intelligent real-time interactions with customers in the web or call center channel. In cross/up-selling, decisionengines execute predictive models reflecting customer behavior. The right customer gets the rightoffer in right time. This boosts revenues as various business cases have shown.

4.5 Analytical Services

In a SOA as infrastructure for BPM, embedded analytics is implemented via analytical services.We can consider analytical services as the technical components of CPM. Analytical services areimplemented as encapsulated, component based modules that can, but need not communicate viaweb services. They publish business logic. This includes all kinds of analytical content andfunctionality, e.g., customizable and extensible templates for metrics in business scorecards andanalytical tools. It also includes all necessary development services for managing the life cycle ofanalytical services (implementing, customizing, maintaining). This is why analytics extends thetraditional data warehouse centric BI. It puts intelligence into the context of strategy, goals andobjectives, processes, and people via metrics and predictive models, and it implements intelligencethrough analytical services.

Reporting, Query and Analysis Services – In a SOA, traditional business intelligence toolsfunctionality for reporting (interactive, production, and financial reporting), querying (ad hocqueries, OLAP) and analysis (data visualization, data mining, statistical tools) is implemented ascomponents providing analytical services that can be embedded in any collaborative process. In aSOA, these services can use any information service for data supply so that these services cannow act on composite data stemming from analytical and operational data sources. Analytics goesreal-time whenever relevant for the business.

Planning and Simulation Services – Planning is a typical cross-departmental process that is bestimplemented as a SOA based process. So, planning functionality is implemented as planning and



simulation services providing full flexibility and adaptability of this process to changing businessscenarios. The advantage of implementing planning through a SOA is obvious, the planningprocess can be composed out of any analytical and other services avoiding the redundancy inanalytical functionality by implementing a planning application in a traditional data warehouse /business intelligence architecture and by fostering a rigorous and audit-proof planning by acontrolled process instead of spreadsheet based manually driven planning processes. (Fig. 9)


Figure 9: Planning and simulation with Cubeware Cockpit: A number of planning approaches can be used,such as top-down with splashing, bottom-up and counterflow. As well as handling several planningversions, both forecasts and "what-if" simulations can be set up. Current actual values can be accessed bythe planner in the same report. The illustration shows the simulation of the profit and loss assuming a 4 percent raise in salaries in relation to the plan IV values of the previous year, taking account of estimated paycost factors over the coming year.

Dash-Board Services – A dashboard (Fig. 10) visualizes large volumes of information stemmingfrom various data sources in a compressed way. Degree of compression and the kind ofvisualization depend on the goal and on the user. A dashboard can also be used to implement abusiness scorecard. It is typically embedded as a portlet in a portal framework. An informationprofile is at the core of a dashboard. It describes which information, functions, knowledge andprocesses an information consumer (employee, customer, supplier, partner, dealer etc.) must haveaccess to according to his / her role. Based on the information profile, the dashboard ispersonalized according to the paradigm of the information supply chain: Each informationconsumer gets exactly what he / she needs to do his / her job. Deployment is either passive, i.e.the information consumer uses search and navigation services to access its metrics and is guided



by an analytical workflow, or active, i.e. in case of escalation, events or alerts, importantinformation is sent to the information consumer by special channels, e.g. SMS, instant message, email etc. for triggering decisions and actions. This enables management by exception.


Figure 10: Example of a Dashboard created with BOARD. BOARD is a CPM toolkit (see chapt. 8) that offersa variety of presentation services for visualizing indicators including gauges, stoplights, thermometers andcockpits. Drill-through and ad-hoc queries for more detail about what the dashboard shows you are otherhighlights as well as self-service dashboard creation and customization.

Data Integration – Traditional business intelligence tools worked on the data warehouse, whereasanalytical services work on a data integration platform. In a SOA, the data warehouse becomes abackend data service, and the data integration platform is part of the enterprise service bus (Fig. 2and 8). Data integration provides information services for analyzing data, master data and metadata, develop data models, prepare and profile any type of data, as well as ETL (extraction,transformation, load) services. For more on data integration see chapter 5.

Interactive Analytics – Embedded analytics is complimented by interactive analytics (dataexploration environment). Metrics are not only derived top-down from strategy, goals, andprocesses, but could also be derived bottom-up from data. This is the purpose of interactiveanalytics. Up to now, mainly traditional business intelligence tools like data mining, statistical tools,adhoc querying, OLAP tools etc are used, but now on top of the data integration platform. In themean time, there is a new generation of analytical tools coming to market that provides ratherenhanced data visualization techniques and analytical workflows.



Tools for interactive analytics are used by interdisciplinary teams. A specialist for tools andmethods and a power user representing the future information consumers jointly drive thisanalytical process. The necessary data services for supplying the tools are provided by an ITspecialist. The IT specialist ideally is a data architect who knows well the enterprise data and datasources and who can identify and evaluate external data sources and services for enriching theinternal data. In this sense, interactive analytics is still a special task for especially trained expertswith specialized tools. But the new generation of analytical tools brings additional improvements.They have added collaborative tools for better team support so that communication andcollaboration between the different parties engaged in the analytical process is sufficientlyenhanced.

Real-Time Analytics – Interactive analytics is a highly interactive process driven by man. Whenthe amount of data to be analyzed is huge (e.g., in the order of terabytes), then the tools becomethe bottleneck, not the interdisciplinary team driving the process. Then, real-time analytics couldhelp. Real-time analytics is based on three different principles that can be also combined. Vendorsare listed in chapter 8.

• Special Database Technologies – Technologies like compression, indexing, vectorprocessing, memory-based caching etc. can dramatically improve the performance of adhocqueries and other business intelligence components / tools. This speeds up the explorationprocess by faster responses (from hours to minutes and seconds). Technologies in thiscategory are rather mature.

• In-Memory Databases – This is one of the rather recent developments in databasetechnology. Here, the total database is processed in memory providing even more performancethan the specialized data base technologies that still store data physically. In-memory databases especially benefit from a 64 bit address space.

• Special algorithms – They are used for reading and processing data. They overcome thelimitations of traditional SQL and OLAP technologies. Many vendors combine these featureswith special data base technologies as described above.



arcplan Analytic Services

Data Sources

arcplan EnterpriseStandard Sales Report

otherRDBMS OLAP

Web Services CommunicationInterface

Integration via Web Services,here: Market shares

ODBC,MDX,XMLA,…1

2

arcplan Enterprise 5Web Services Communication Interface: external to arcplan

Internet Services

BusinessApplications

BusinessProcesses

ExternalInformationaccessible viaWeb Services

11

arcplan Analytic Services

arcplan Enterprise Report

N.N.

Desktop User / Workplace

arcplan Analytic Services Provider

arcplan Enterprise 5arcplan Analytic Services Provider: arcplan to external

511

Figure 11: Example for embedding analytical services (here via web services). arcplan Enterprise consumesweb services that are orchestrated and presented by its analytical workflow (top). arcplan Enterprise alsopublishes services via web services that can be orchestrated by other services and processes (bottom).



5 Data IntegrationData integration became an issue, when it was about to fill and to refresh a data warehouse.Solutions have been extraction, transformation and load (ETL) processes. But in the times ofprocess-orientation and CPM, data integration gained a much higher, even mission-criticalimportance, and a much more wide-spread usage.

5.1 Data Integration Platform

It has been common practice to supply a data warehouse by ETL processes. ETL processes areeither supported by batch and / or message / queuing, depending on whether time is critical fordata supply. This will continue, and this is one task that is still addressed by data integrationplatforms. But now we need more. We need information and data services (Fig. 8) enabling thesimultaneous access of data warehouse and operational data via a data integration hub (Fig. 2). Inthe past, one has tried to solve this time critical data access problem via an ODS (operational datastore). Using the ODS approach is not always sufficient, because storing data in an ODS mayalready exceed a given time window, and unfortunately, business logic needed for calculating morecomplex metrics may be hidden in the application logic and is not available on the data level.

There are two options: low latency and zero latency data integration. So, the key point is first todetermine what latency can be tolerated for a given process. Note that latency is correlated withcost: the lower the tolerated latency, the higher the cost.

The low latency model is based on a data integration platform that collects all relevanttransactional and analytical data and stores it in a so-called low-latency data mart (LLDM). Thisrequires integration of the data integration platform with the integration hub (ESB) where theprocesses and services across all backend applications are managed. The LLDM is refreshedeither by message queuing or by batch, where the batch is executed in short periodicitiesaccording to the tolerated latency (e.g., hourly etc.). The LLDM can be used for low latency datapropagation. This is a feedback loop for triggering events in operational systems via cross-processmetrics. This coupling with operational systems requires managing the data integration platformlike the ESB: The data integration platform is an operational system. (Fig. 12)

This model is different from an operational data store (ODS) where data from operational databases is stored via ETL processes. So, all transaction logic that is not stored in the operationaldata bases cannot be mapped to operational data stores. Furthermore, the ETL process is notsynchronized with the transactions, i.e. ODS data is not always in synch with the state oftransactions. This stresses the need for low latency data marts, especially in the case of legacysystems.

The zero latency modell is also called Enterprise Information Integration (EII). It can beunderstood as a logical data base access layer spanning across all operational data bases and thedata warehouse providing information and data as services. The access is done via XML and the



EII resolves the data request into various SQL statements accessing the corresponding data basesand transforming the data so that the requested compound data is published as a service andavailable for the process. Indeed, such an information service could be also implemented as a webservice.


Data LatencyData Latency

OLTPOLTPSystemSystem







Data Integration PlatformData Integration Platform LLDM

Enterprise Information Integration (EII)Enterprise Information Integration (EII)

Real-TimeAnalytics

Real-TimeAnalytics

Real-TimeData Propagation

Data or Events

„zero-latency“

„low-latency“

Figure 12: Real-time data integration can be implemented as a low latency or zero latency solution. The lowlatency solution works with a low latency data mart (LLDM) that stores transaction synchronous, but bylatency time delayed information and data for BAM/PPM. The zero latency solution (EII) means transactionsynchronous access to heterogeneous OLTP data (OLTP = online transaction processing). Real time datapropagation triggers operational systems with events based on cross-process metrics via a data integrationplatform.

5.2 Information Services

We already have met information services as a special category of SOA services (Fig. 8). Note thatinformation services do not only make perfectly sense in a SOA, but bring also great value toenterprises that are not planning to use a SOA as their IT architecture, but suffer from datafragmentation. (Data is kept in data silos in isolated applications or data marts). Let us start with amore detailed definition of an information service.

An information service is a modular, reusable, well-defined, business-relevant service thatenables the access, integration and right-time delivery of structured and unstructured, internal orexternal data throughout the enterprise and across corporate firewalls. An information service canbe a meta data service, a master data service, or a data service.




Infr

astr

uctu

reSe

rvic

es

Met

a an

d M

aste

rD

ata

Serv

ices

Architecture of Information ServicesArchitecture of Information Services

Delivery Services

Data Integration Services

Universal Data Access ServicesAdm

inis

trat

ion

Serv

ices

Web ServicesAnalytical Services

Portals

INTERNAL EXTERNALInformation CONSUMER

Applications ApplicationsProcesses

ApplicationsDatabases

MessagesFlat Files

XMLunstructured Data

Mainframe

Data Sources

JMS Web Svc SQL JDBC WebSvc

Figure 13. To overcome data fragmentation, information and data is delivered by information services.Information services include six different categories. The architecture is shown in the figure. Universal dataaccess services provide access to any internal or external data source. Data integration services provide anytype of mapping, matching, and transformation. Delivery services publish information to any informationconsumer – internal or external. Meta and master data services provide the common business vocabulary.Infrastructure services look to authentication and security. Administration services provide the functionalityfor administrators, business analysts and developers for managing the life cycle of all services.

Given the definition of an information service, the next step is now to look at the needs ofinformation service consumers to identify the different categories of information services and theirarchitecture. (Fig. 13)

• Universal Data Access Services. Access services are the basic CRUD services for creating,reading, updating and deleting data from any backend systems, structured or unstructured,internal or external. Access services also provide zero and/or low latency access to federateddata. This is sometimes called enterprise information integration (EII).

• Infrastructure Services. Infrastructure services include basic functionality aroundauthentication, access control, logging, etc.

• Data Integration Services. Integration services move data from source data models to targetdata models like synchronization, transformation, matching, cleansing, profiling, enrichment,federation, etc.

• Meta and Master Data Services. Their purpose is to manage and use the technical andbusiness metadata and master data for audit, lineage, and impact analysis purposes.



• Data Delivery Services. They automate and standardize the publication of information to allconsumers according to a request/reply model or a publish/subscribe model (data syndication).Delivery mechanisms are bulk and/or single records by batch, real-time messaging or deltamechanisms bases on change of data.

• Administration Services. These are services for the life cycle management of the otherservices, i.e. development, management, and monitoring and controlling.

The model of service-orientation provides another advantage. Due to the sub-service principle,composite information services can be built for any purposes. Typically examples consist of datawarehousing, data migration, and data consolidation processes. We will discuss other exampleslike master data management and data quality management in the following chapters.

5.3 Meta and Master Data Management

Process-orientation needs meta data management. The meta data layer spans all layers of theSOA. Meta data is key to a consistent data model including life cycle management for a consistentcomprehension and communication of the data model, for data quality, data protection andsecurity. Meta data builds the business vocabulary of the enterprise and even across enterprises.Meta data is organized by three layers:

• Layer 1 – Master Data: This is business oriented meta data providing the foundation of thebusiness vocabulary. Master data is meta data describing business structures like assets,products and services, and the business constituents (e.g., suppliers, customers, employees,partners etc.) This provides the famous single view on all enterprise structures.

• Layer 2 – Navigational Meta Data: Meta data on navigation (e.g., sources and targets of data,cross references, time stamps)

• Layer 3 – Administrational Meta Data: Meta data on administration (information profilesincluding responsibility, security, monitoring and controlling usage etc.)

Meta and master data provide the single point of truth that was traditionally claimed by the datawarehouse. Today, this single point of truth is established through data integration. The businessvocabulary plays the central role. It controls both, BPM and CPM: Processes and metrics need acommon and uniquely defined language for modeling and for communication to all businessconstituents in collaboration contexts. Master data describes the data objects of processes andrules: No processes, metrics, and rules without data.

Master data in a SOA is provided by special information services. Traditionally, master data wasapplication dependent, and it was re-implemented in always new versions together with thecreation of new applications.

Example. When master data is fragmented across various application islands, then eachapplication tends to develop its own terminology. Lack of consistency directs to chaos.Product and order numbers in one application do not match with those in other applications.



Collaboration with suppliers and customers gets more and more expensive. Each time, anew customer, a new supplier, a new product is introduced, a lengthy, cumbersome, anderror prone procedure adds the new items to translation tables, or even new translationtables have to be created. Changes get slow, introduce quality problems, and boost costs.But when master data is provided as an information service, then one simple update in themaster database propagates changes safely and automatically to all related systems.

The example emphasizes again the problems of application orientation. Applications are like silos.Terminology and models end at the border lines of an application. Cross application processes areinterrupted, metrics, rules and business vocabulary are inconsistent and redundant. Efforts tointegrate and synchronize across application islands get more and more complex and expensive.IT gets stuck into the maintenance and becomes a blocker to the business. Agility is not feasible inthe traditional world of applications. Process- and service-orientation is the way out. Applicationindependent processes, metrics, rules, and master data are the pre-requisite for agility.


Meta and Master DataMeta and Master Data

Operational Data

Meta Data (DNA)

Analytical Master Data

Transactional DataOperational Master Data

l Customerl Partnerl Supplierl Productsl People

l Customerl Partnerl Supplierl Productsl People

l Spacel Timel Planl Organization

l Spacel Timel Planl Organization

Figure 14: We derive master data from operational data (of the OLTP – online transaction processing –systems) and we classify master data into operational and analytical master data. Master data is part of thebusiness layer of meta data (D = Definition, N = Navigation, A = Administration).

The services repository provides a container of all meta and master data. It plays the role of theintegration hub for the meta data of all back end systems in the BPM model. In a SOA, whenservices of back end systems are invoked by an integrated solution representing the integrationlogic and the flow of the process, then they must speak to each other in the same language that isbased on the business vocabulary of the repository. A point-to-point communication would again



lead into the chaos of isolated islands. The only solution is to transform the meta data model ofeach back end system into the central business vocabulary of the repository of the BPM integrationhub. Then, all back end systems can speak to each other and adding additional back end systemsbecomes straight forward, easy, and fast.

A subset of layer 1 of meta data is used to describe master data. Master data describes thestructures of an enterprise (Fig. 14). There is operational and analytical master data. Operationalmaster data is part of transaction data, where transaction data consists of (operational) masterdata and inventory data. The various types of operational master data can be derived from thebasic structure of an enterprise. These are all objects and people that are involved in executingand managing business processes, i.e. products and business constituents (customer, supplier,dealer, employees). Analytical master data can be derived from CPM and the process ownershipmodel. It represents the principles of measuring and responsibilities: time, space, plan andorganizational units (e.g., cost center, cost objective).


Transparency and TraceabilityTransparency and Traceability

Supplier CustomerEnterprise

Collaborative Processes need aCommon Business Vocabulary.

Master Data Management (MDM) – Pre-Requisite forTransparency and Traceability

MasterDataSynchronizing Versioning

The 3 pillarsof MDM:§ Data Integration§ Data Profiling§ Data Quality

Figure 15: Master data management is all about to establish information services for synchronizing andversioning of all master data across all backend applications with a SOA. The center of master datamanagement is a repository that manages the common business vocabulary so that all processes can useunique structures and terms. Best practice architecture for such a repository is a hub & spoke architecturecorresponding to the architecture of an ESB. The three pillars of MDM are discussed in chapt. 5.4

Meta data and master data are not static. On the contrary, any merger and acquisition, any marketchange, any internal organizational restructuring, any update of a business definition and rulecreates new meta data and master data. But it is absolutely insufficient just to update meta data



and master data and store the most recent and actual version in the repository or in a data base.For enterprise planning and for any comparisons between past, now, and future, theavailability of the total life cycle of all meta data and master data is a must. This is why metadata management and master data management are to be based on life cycle management. Therepository must include the life cycle of all meta and master data. (Fig. 15) Today, this is a weakpoint, sometimes even a gap in vendor offerings and enterprise architectures. But nevertheless,without meta and master data management, BPM, CPM and SOA initiatives will fail.

5.4 Data Quality

What is the day in the year when most people have their birthday according to all birthday datastored in all data bases in the world? Nonsense question? Not at all. The result is striking. It is the11th of November. Why? Well, if a new customer is to be entered into the customer data base, thenthere are mandatory fields to be filled in and additional fields. Input into mandatory fields ischecked (in many situations at least), but input into additional fields is typically not checked.Birthday data unfortunately is stored in additional fields. So what happens? Man is lazy, and theeasiest and fastest way to input a birthday date is “1-1-1-1”….

Enterprises have introduced ERP systems from SAP and others for many millions of euros. One ofthe drivers was to be more competitive based on all stored data about market and customers.CRM per self-service, coupons, pay-cards, communities, and weblogs are best practices forchasing the budgets of customers. Customer-orientation is the rule. Marketing, sales and serviceare working together supported by collaborative end-to-end processes. Inbound and outboundcampaigns in customer interaction centers and/or in the web shops are enriched through businessanalytics. The demand and customer driven supply chain is getting more and more reality. But aswe may have already noticed: Data quality is the prerequisite.

Example. In the apparel industry, people are already used to collect all sales transaction datain a data warehouse. Customer profiles are calculated, and a demand profile per boutique canbe derived. According to these demand profiles, merchandise is individually attributed to allboutiques. As a consequence, customer will typically find “his/her” products he/she is lookingfor in “his/her” boutique. Customer satisfaction and loyalty increases: In the end, customerprofitability increases. There is another consequence: We also cut costs. If a boutique offersthe right products to the right customers, then stock is lowered, and lower stock means lesscosts. An economy of 30% to 40% of cost of stock is achievable.

Data quality is key for being more successful with information. The principle “garbage in – garbageout” is without mercy. When enterprises do not notice before building CPM solutions that the datastored in SAP and other backend systems is insufficient for SOA based processes, then it is toolate!

Example. A leading European mail-order-house had a problem with its birthday data. Birthdaydata allows the calculation of the age of customer, an important parameter in customerrelationship management. So, what can you do, if your birthday data is not reliable? There is a



solution that provides a good estimate of age of customer. You look to the customer’s firstname. First names follow trends. Customer age can be estimated based on patterns ofattributing first names to children. But this is an expensive approach, and it will never achievefull reliability. Much better approach is to build data quality from the beginning into theoperational processes.

Building quality from the very beginning into processes – that is a well known concept. Indeed, thisis the idea of “total quality management (TQM)” that has been applied successfully inmanufacturing 20/25 years ago. TQM for IT, this is not only an issue for today. When implementingERP systems, the principles of TQM for assuring data quality should have been already applied.But still today, data quality is an issue. In many enterprises, a data quality director and sponsor onthe executive level is the exception. Data quality needs management attention.

Example. Let us assume, we want to build a 360° view on customer. Goal is to knowcustomer in order to optimally serve customer according to his/her customer value. 60% to80% of cost for customer data integration is caused by infrastructure. Customer dataintegration means to synchronize and to versionize customer data from various sources intoone single customer data model. Data stems from various application islands, historical andarchived data bases, external market data, demographical data, web click stream data andothers. When building the customer data model, you may notice at once that in a backendsystem, there is a data table with data about customer that could be linked to a table inanother system providing a new and not yet available customer insight. Great, but what if thedata field that is to be used as key is not a mandatory field, but just an additional field?Typically, this is the end of the good idea: Will the owner of this application be ready to changethe additional data field into a mandatory data field just because you tell him that would easeyour job? An IT question turns into a business issue. Indeed, only the business can decide onthese questions that seem to be IT questions at first glance, but have to be tackled and solvedin a collaborative approach jointly by business and IT.

Data quality needs management attention as this example proofs. Leading enterprises have dataquality directors that report directly to the CIO. The CIO brings data quality to managementattention and creates a change culture task. The data quality director coordinates the roles of datastewards and data custodians. Data custodians are located in the lines of business. The have toresponsibility for data quality from the business point of view, i.e. content of master and meta dataas well as validity rules for certain transaction data. In process-oriented enterprises, the role of adata custodian can be played by the process owner. Data stewards are associated to the datacustodians. Their role is to implement all rules and models within the IT systems. Their skill setshould include data base administration and/or data administration.

Data quality should be implemented into operational processes by a TQM program. The principlesare: Preventing is better than healing.

But what if it is already too late for prevention? What can be done to improve data quality inalready existing data bases? There are two complimentary types of tools:



Data Profiling. Data Profiling is used to analyze the properties of a given data set and to create aprofile. There are three types of analysis:

• Column profiles. Analysis of content and structure of data attributes helps to identify dataquality problems related to data types, values, distributions and variances.

• Dependency profiles for identifying intra-table dependencies. Dependency profiling is related tothe normalization of a data source. It provides expected, unexpected and weak functionaldependencies and potential key attributes.

• Redundancy Profiling. It identifies overlapping inbetween attributes of different relationships.This is typically used to identify candidate foreign keys within tables, and to identify areas ofdata redundancy.

Tools for data profiling use methods from descriptive statistics (analysis of distributions, tests foroutliners) as well as data mining (cluster analysis and decision trees). Data Profiling provides ananalysis “as is”, and is a valuable tool for estimating and directing further investments in dataquality. Data profiling tools identify data quality problems much faster than any manual analysis.

Data Cleansing is based on different methods:

• Parsing. Compound data is decomposed.

• Semantic approach. Data is transformed into standard values and formats according to rules.

• Benchmarking. Internal data sources are compared with external sources for verification.

• Matching. Data of similar content in different fields is identified (match customer informationthat is stored in different applications to one and the same customer).

• Removing duplicates. (e.g., address data)

• Consolidating. Create a complete data record out of dispersed information (e.g., create onecustomer address record)

• House holding. Detect relationships between data (e.g., identify all persons belonging to oneand the same household).

• Enriching. External data may enrich the value of cleansed enterprise data.

Data cleansing tools are based on probabilistic, deterministic and knowledge procedures.Probabilistic and deterministic procedures use appropriate algorithms, whereas the knowledgebased approach uses country/language specific knowledge data bases for composing addresses,names or legal entities.

Managing data quality standards again is a process. It combines proofing and cleansing activities:Both provide information about the status of quality progress. A periodic execution of this processhelps to continuously monitor and control enterprise data quality. It should be part of the TQMmodel. This process as all processes should be supported by a BPM tool and run on a SOA. Thelogic of the TQM model orchestrates the data quality services of profiling and cleansing.



6 Latency mattersToday, CPM must address the operational, tactical, and strategic aspects in a seamless way.Leading process-oriented businesses use highly automated processes for straight throughprocessing. Metrics trigger decision engines and actions are taken in an automated way. Just incase of exceptions, escalation management, authorization, entry of triggers (self-service), andwhen applying collaborative services human interactions are (still) needed. Now, when theidentification of alerts and exceptions becomes time critical, human interactions even become toslow. This is where latency matters and action time becomes critical (Fig. 16). The action timemodel shows three critical phases, data latency, analysis latency, and decision latency.


Real-Time and Action-TimeReal-Time and Action-Time

Event

BAM

Info

rmat

ion

Stor

ed

DecisionEngines

DataLatency

AnalysisLatency

DecisionLatency

Value

Time

Action Taken

DataStored

Real-TimeData Integration

Action Time

After: Richard Hackathorn and Colin White

Figure 16: In operational CPM (BAM), time may be critical. The action time model decomposes action timeinto data latency, analysis latency, and decision latency, and it shows by which approaches, action time canbe minimized.

Data Latency. This is addressed by real-time data integration (Fig. 12). There are two options aswe have already seen, low latency and zero latency data integration. For a discussion, we referback to chapter 5.1.

Analysis Latency. This is addressed by BAM solutions: Analytics must be now available in realtime. Since analysis latency depends on the complexity of events, we first discuss the varioustypes of events in order to understand the different kinds of BAM solutions and their constraints toanalysis latency. We follow the approach taken by Luckham (2002).



• Simple events. These are events where all data necessary to detect the event are availablewhen the event happens. We have already seen examples like product availability anddeliverability. The goal of BAM now is to compare such an indicator with a threshold to launchactions for control. Here, the really critical part of latency is mainly data latency since analysislatency caused by the calculation of the indicator is rather small compared to data latency.

But analysis latency becomes an issue when using predictive models. In many situations, thepredictive model cannot be derived in real time – data mining does not work in real time. This isthe reason why modeling of predictive models by data mining processes was strictly separatedfrom applying predictive models in operational processes. So, usage of a predictive model isreal-time, but not modeling. The predictive model was exploited in an off-line mode with thehope that the model that is based on the past maps the actual and future. An approach toovercome this problem was periodically remodeling the predictive model with the speed ofsupposed changes (e.g., week, month). New approaches and technologies make a breakthrough. Predictive models can be made self-learning by adaptive algorithms. They matchdynamically to the changing process context. Such an adaptive predictive model is always on-line and maps to the presence based on the actual data driving the adaptive algorithm. Thispresents a low latency solution for analysis latency. In fractions of seconds, adaptive modelscan be recalculated. This enables an application of adaptive, dynamic models for intelligentcustomer interactions in call centers or in web shops.

• Event streams. This is a continuous time sequence of events. For example, timingcorresponds to the arrival times of events in a BAM tool or by time stamps. Monitoring andcontrolling of traffic in all kinds of networks is typically based on event streams. Examples canbe found in telecommunications, information processing, air traffic, ground traffic etc. For BAMtools, there are different application domains to be distinguished.

o Simple pattern recognition. BAM tools for this type of problem are based on time seriesanalysis. The goal is the forecast, i.e. the prediction of the outcome of the next event.Typical examples are sales forecast as well as forecast of stock prices or peaks inconsumption.

o Complex pattern recognition. Events streams can be conditional. They could happen atdifferent locations at different times and influence each other. BAM tools are now basedon multivariate time series analysis. Examples are concurrent and collaborative processeslike sales promotions of several competitors in one and the same market. Task for a BAMtool could be to track the effectiveness of its own marketing activities, to measure theimpact of its promotions and to derive marketing strategies for defense or attack based onthe BAM.

o Pattern abstraction. Subsequent events could be detailed events of an event on a higherabstraction level. BAM tools are now used to detect and identify the typically higher valueabstract event based on the evaluation of the single and isolated detailed events persemantic reasoning. For example, consider the analysis of buying signals of a customer.Customers send signals about their readiness to buy a certain product, in particular if the



investment exceeds a certain level like buying a car or a house. A BAM tool should nowdetect buying readiness as soon as possible given received buying signals so that salesget a window of opportunity towards a competitor.

BAM tools for event streams are based on special fast algorithms (e.g., matching algorithms andother semantic methods). Besides the rather well known domain of time series analysis, this is stilla young area of development and many solutions are in an experimental stage.


Figure 17: Today’s tools for data analysis and data mining like the STATISTICA Data Miner support a varietyof statistical methods for developing data driven decision models and comparing their prognostic capabilities.After selecting the appropriate predictive model or set of models, these tools automatically apply the rules asshown in the example of customer scoring.

Decision Latency. Indeed, when time matters, decisions cannot be taken anymore by humans.Now, decision taking must be automated by decision engines. Decision engines are based on ruleengines. Rules can be generated bottom-up via predictive models. Such a set of rules can berather complex. For instance in e-commerce, intelligent customer interactions use predictivemodels that are derived from various data sources like real-time and historical surfing properties,buying patterns, busing history, catalogue information, sales strategy, and other external conditionslike time of the day, day of the week, and seasonal information to get recommendations with highrelevance. In many cases, the set of rules is simplified and reduced to one single parameter, ascore. (Fig. 17). Decision rules and scores are identified by using various data sources, andpreviously detected data structures and patterns.

Rules may be also specified by experts in a top-down approach. This is a certain revival of the oldexpert systems popular in the late 80s and early 90s. Ultimately, rules engines can be modeled bya combination of predictive models with expert rules. Decision engines have been discussed indetail in Martin (2003-B).



7 CPM and classical BI: fundamental differencesCPM has evolved from the old decision support and business intelligence approaches, but today,CPM is a completely different model then classical business intelligence.

• CPM is a top-down model that begins with business strategy. Business Process Managementlinks process analysis and design with cross-functional and cross-departmental process flowsand CPM. Process performance metrics are created at the same time as the processes.

o Business Intelligence (BI) was bottom-up and not process-oriented.

• CPM is based on an information supply chain model that permanently synchronizes theprovision of information with the need for it.

o Business Intelligence was solely an information-providing model (Bill Inmon"Information Factory").

• CPM is a closed-loop model that controls and monitors business processes at operational,tactical and strategic levels.

o Business Intelligence only supported decision making, but not action taking. Theoperational aspects of Business Intelligence were not covered by a coherentapproach.

• CPM metrics are forward-looking. Predictive models enable the identification of problemsbefore they appear. Of course all traditional retrospective metrics remain useful.

o Business Intelligence was retrospective (based on the past). The focus was onanalysis and diagnosis. Potentials of predictive models were not exploited.

• CPM enables transparency by means of the sharing and filtering of information in accordancewith the process ownership model. Everybody gets the information required in the context ofhis/her processes.

o Business Intelligence tools did not provide the information consumer with sufficientinformation. Either one had information that was not accessible (on occasion evenhidden or held back), or you had an absolute flood of data ("information for themasses"). This spoiled acceptance.

• CPM is based on analytical services that are published, consumed, and orchestrated in thecontext of a SOA (service oriented architecture).

o Business Intelligence was a tool-related approach, based on proprietarytechnologies. This resulted in stove piped information silos.



8 Players in the CPM/BI Market

During recent years, the BI market has shown an over average performance with a two digit growthrate. We believe that the following three statements deliver good insight and good argumentsabout reasons: the extensive rearrangement of the BI market, the evolution of BI role, methods andtools, and the transition to CPM.

Statement 1: The market for Business Intelligence disintegrated. For several years already,we have observed increasing merger and acquisition activities in the market. The clou happened in2007 when three mega acquisitions took place: Oracle/Hyperion, SAP/Business Objects andIBM/Cognos. No real big independent BI vendor exists anymore (only exception: privately heldSAS). In consequence, there is no independent BI market any more, but it has been absorbed bythe BPM/SOA, respectively ERP II market. Indeed, the big four in the BPM/SOA market are allleading BI vendors, and this holds also for the ERP II market. The market leader Infor givesalready a good example. This did not happen by surprise, we anticipated this trend already in 2006(expert opinion in is-report 3/2006). But in the new, extended market, the remaining small andindependent BI vendors can very nicely occupy interesting and lucrative niches. The on-goingprocess and service orientation (including SaaS, the new license model for consuming externalservices in a SOA) empowers a best of breed model for vendor selection more than ever, becauseintegration is no more a challenge, but is a given. As a consequence, the outlook for the smallerplayers is excellent due to this market move. Additionally, new markets for “intelligence” spun offthe former BI market, for instance, content intelligence, customer intelligence, financial intelligence,and competitive intelligence. These emerging markets offer new growth opportunities for newand/or repositioned vendors. Indeed, the disintegration of the traditional BI market does not meanthe dead of the market, but a restart with many opportunities for all players.

Statement 2: For enterprises, BI is more important than ever. During recent years, BI changedand evolved a lot. BI became operational, BI was finally put into the context of business processes,and BI was extended to a closed loop model for monitoring and controlling business processes. Inthe end, the old paradigm that BI only works on top of a data warehouse was shown to be toorestrictive and insufficient. Operational data sources became equal to classical data warehousedata. The data warehouse stopped to be the single point of truth. A new challenge was created.Traditional ETL processes continue to be necessary, but we need more to end up in a trueinformation management. The idea of enterprise information management was born. It alsoenables the transition from BI to CPM. And today, we even move further. The key role of BI forCPM extends to governance, risk management, and compliance (GRC). BI enters the board level.

Statement 3: BI arrived at C level. Indeed, CPM and GRC are within key responsibilities of theboard. In the past, BI was sold to the IT. Many BI projects simply suffered due to this ill position.The creation of value by BI was very difficult to show, sometimes even abnegated. Huge datawarehouses caused costs, but nobody really liked to use all the existing data. This is nowchanging. The value creation of BI in the context of business processes is beyond dispute. GRC



turns the office of the CFO to the control room of the enterprise. Even the role of the CFO ischanging. The CFO undergoes a metamorphosis to a CPO, the chief performance officer with fullresponsibility for GRC.


Taxonomy CPM/BITaxonomy CPM/BI

BAMBAMBAM

Decision Engines

Data Integration PlatformData Integration PlatformDataData

WarehouseWarehouse

LLDMLLDM

ETLETL

CPMCPMCPM

ActionAction

BusinessScorecard

Enterprise Service BusEnterprise Service Bus

Figure 18: Action time (see Fig. 16) based taxonomy of CPM market players. (BAM = business activitymonitoring, ETL = extraction, transformation, load; LLDM = low latency data mart, EAI = enterpriseapplication integration, MQ = message / queuing)

Let us now move to the market players. From the three phases of action time (Fig. 16), we canderive a taxonomy for classifying the players (vendors) in the market (Fig. 18). Key players in thedifferent categories are listed below, but we do not claim to provide an exhaustive list. More detailson specific vendors will be published in part 2 of this white paper, where in each white paper wewill map the vendors’ architecture and strategy to the vision and reference architecture developedin this part 1.

Part 2 – Available white papers (February 2008): arcplan, Cubeware, EPOQ, Informatica,in-factory, Panoratio, SAP, Spotfire (see www.wolfgang-martin-team.net) A white paper on Lixto isplanned for April 2008.

CPM Toolkits provide the actual state-of-the-art implementation of CPM. The term goes back toGartner Group. CPM Toolkits are positioned between CPM Suites and single tools (likespreadsheets). CPM Toolkits provide CPM specific functionality as services: They are open andhave standardized interfaces, whereas CPM Suites are built on proprietary technologies and aretightly integrated. This makes implementation of CPM Suites difficult: It is not easy to customizethem, and there needs rather high efforts and costs to integrate a CPM Suite into existing

http://www.wolfgang-martin-team.net)



applications. CPM Toolkits follow the SOA principles. This loosely coupling of CPM services hasseveral advantages. Service-orientation eases up customization. Services can be used like LEGObuilding blocks. They are easily invoked by right mouse clicks and are ideal for mash ups. The enduser can build its own composite application: By mashing up CPM services, the user creates theanalytical workflow and orchestrates the analytical services provided by the CPM Toolkit. This iswhy CPM Toolkits address the old requirement that BI tools should enable the business to workautonomously and to build their own reports and analyses without programming.

The following listing of vendors is not supposed to provide a complete view on the market. But it isquite comprehensive and puts a focus on the German speaking markets: It includes quite anumber of local players.

Data Warehouse / Data Mart (data base technologies – MOLAP, ROLAP, relational & specialtechnologies))

BOARD, Computer Associates/Cleverpath, IBM, IBM/Cognos/Applix, Infor/MIS AG, instantOLAP,InterSystems, Kognitio, KxSystems, Microsoft, MicroStrategy, MIK, Netezza, NCR/Teradata,Oracle, Panoratio, Sand Technology, SAP, SAS, Sybase, Xcelerix

Open Source: The Bee Project, Greenplum, Jedox, mySQL2, Pentaho

Business Intelligence / Business Analytics (Frontend: Suites, Toolkits and SpecializedTools)

§ The BIG FOUR: IBM, Microsoft, Oracle, SAP

§ Leading world-wide specialists: Actuate, Information Builders, MicroStrategy, SAS, SPSS

§ Challenders: arcplan, aruba Informatik, Aruna, BOARD, CA/Cleverpath, Cubeware, Group 1,Human IT (InfoZoom), Infor/Extensity, instantOLAP, Menta, MIK, Panorama, Panoratio,Prevero, QlikTech, Q4bis, SAMAC (nur für IBM iSeries), Targit, Tibco/Spotfire

Open Source: Actuate/BIRT, The Bee Project, JasperSoft, Pentaho

Business Activity Management/Complex Event Processing (BAM/CEP)

Aleri, AptSoft, Axway, Business Code, Coral8, Coremetrics, EPOQ, Gemstone, HP TSG, IBM, IDSScheer, Information Builders, Microsoft, Oracle, Senactive, SL Corporation, SoftwareAG/WebMethods, SUN/SeeBeyond, Systar, Tibco, Vitria, WebTrends

Business Scorecards

Active Strategy, Actuate/PerformanceSoft, Antares, arcplan, aruba Informatik, BOARD, BOC,Business CoDe, Coda, Communic, Corporate Planning, Cubeware, Dialog Strategy, hfp, HologramBI, Horvath & Partner, Hyperspace, IBM, IBM/Cognos, IDS Scheer, iGrafix, Infor/MIS, macssoftware, Microsoft/ProClarity, MIK, Oracle, Panorama, Prevero, Procos, Prodacapo, QPRSoftware, SAP/Business Objects, SAS, Stratsys, Targit

2 In January, Sun Microsystems has announced its intension to acquir mySQL.



Predictive Models (Data Mining & related vendors)

Anderson Analytics, Angoss, Autonomy/NCorp, Chordiant, Cognos, Coremetrics, EPOQ, FairIsaac, IBM, Infor/E.piphany Insightful (Miner, S-Plus), ISoft, Kana, KXEN, Magnify, Megaputer,Microsoft, NCR/Teradata, Oracle, Portrait Software, Prudsys, SAP, SAS, SPSS, StatSoft,thinkAnalytics, Treparel, Unica, Verix, Viscovery

Open Source: Knime, Orange, RapidMiner, Rattle, R-Project, Weka

Text Analytics

Aero Text, Anderson Analytics, Attensity, Basis Technology, Clarabridge, Clear Forest, LexisNexis,Linguamatics, Nstein, SAP/Business Objects, SAS, SPSS, StatSoft, Teezir, Temis Group, Treparel

Open Source: Gate, RapidMiner

Decision Engines

Chordiant, Corticon, EPOQ, Fair Isaac, Haley, IBM, ILog, Infor/E.piphany Innovations, Kana,MicroStrategy, Oracle, Portrait Software, Prudsys, SAP, SAS, SPSS, StatSoft, thinkAnalytics,Tibco, Versata, Viscovery

Financial Performance Management (Budgeting, Planning, Financial Consolidation etc.)

Acorn System, arcplan, BOARD, Coda, Complan & Partner, Corporate Planning, Cubus AG,Denzhorn, Hologram BI, IBM/Cognos, IDL, Infor/MIS AG, Longview, LucaNet, macs Software,Microsoft, MIK, Oracle/Hyperion, Orbis AG, Prevero, Procos, Prodacapo, PST, SAP/BusinessObjects, SAP/OutlookSoft, SAS, Software4You, Targit, Winterheller

Data Integration – Platforms

Adeptia, Composite Software, ETI, Gemstone, IBM, Informatica, Information Builders, ISoft,Oracle, Pervasive, Red Hat/MetaMatrix, SAS/DataFlux, SAP/Business Objects, Software AG,SterlingCommerce, Tibco

Data Integration – Special Tools, Webintegration

Qitera, Kapow Technologies, Lixto, Teezir

ETL

AbInitio, BOARD, Cubeware, ETI, Group 1, IBM, Informatica, Information Builders, ISoft, Kognitio,Menta, Microsoft, Open Text, Oracle, Pervasive, SAP, SAS, Sybase/Solonde

Open Source: The Bee Project, CloverETL, Enhydra Octopus, KETL, Pentaho/Kettl, Talend

Data Quality

Address Solutions, Datras, ETI, Group1, Harte Henks, Human Inference, IBM, Informatica,Innovative System, Nokia/Identity Systems, Omikron, Oracle, SAP/Business Objects,SAS/DataFlux, tekko, Uniserv



9 Summary

We believe that our vision on corporate performance management across operations, tactics, andstrategies will become a standard for continuously evolving metrics-driven management. We alsobelieve that our reference architecture of analytical services infrastructures will be a standard fordynamic enterprise specific service oriented architectures (SOA). This Whitepaper will help tomake strategic decisions on strategies and platforms.

CPM is the answer to today’s challenges running a business: You can only manage what you canmeasure. This is one of the leitmotivs that will lead enterprises into a successful future.

Munich, February 2008 Annecy, February 2008

E-Mail-Addresses:

[email protected]

[email protected]

Literature:

Inmon, W.H., Imhoff, C., and Sousa, R.: Corporate Information Factory, New York, John Wiley & Sons, 1998,274 pages

Luckman, D.: The Power of Events: An Introduction to Complex Event Processing in Distributed EnterpriseSystems, Boston, Addison Wesley Professional, 2002, 400 pages

Martin, W.: Business Performance Management und Real-Time Enterprise – auf dem Weg zur InformationDemocracy, Strategic Bulletin, IT Research, www.it-research.net, Sauerlach bei München, 2003-A, 32 pages

Martin, W.: CRM 2004 – Kundenbeziehungsmanagement im Echtzeitunternehmen, Strategic Bulletin, ITResearch, www.it-research.net, Sauerlach bei München, 2003-B, 32 pages

Martin, W.: BI 2004 – Business Intelligence trifft Business Integration, Strategic Bulletin, IT Research,www.it-research.net, Sauerlach bei München, 2004, 32 pages

Martin, W.: SOA 2008 – SOA basierendes Geschäftsprozessmanagement, Strategic Bulletin, IT-Verlag fürInformationstechnik GmbH, Sauerlach, 2007, 28 Seiten

Martin, W., Nußdorfer, R.: Role of Portals in a service oriented architecture (SOA) – “Status and Trend –Processes and People – Presentation and Collaboration Services“, iBond White Paper Vol. 4, www.soa-forum.net, Munich, 2006, 33 pages

Nußdorfer, R., Martin, W.: RTE – Real-Time Oriented IT Architecture: All Together Now, Strategic Planningof IT Architectures, iBonD White Paper Vol. 1, www.soa-forum.net; 2003, Munich, 35 pages

Nußdorfer, R., Martin, W.: BPM – Business Process Management – Änderung des Entwicklungsparadigmas,Kompendium “Geschäftsprozesse als Lösungen“, iBonD White Paper Vol. 3, www.soa-forum.net; 2007, 43pages

mailto:[email protected]

mailto:[email protected]

http://www.it-research.net,



http://www.soa-forum.net;

http://www.soa-forum.net;



10 The Sponsors

BOARD InternationalFounded in 1995, BOARD is a global leader in the BI and CPM (Corporate Performance Management)Toolkit space and offers a unique combination of speed and simplicity. BOARD has enabled over 2000companies worldwide to rapidly deploy BI & CPM applications in a single integrated environment completelyprogramming free and in a fraction of the time and cost associated with traditional solutions.

BOARD provides one accurate, corporate view of your information, fully integrated with your CPM processesand therefore uniquely linking performance to strategic vision at all levels down to operational detail. A toolkitapproach to Corporate Performance Management (CPM) represents a new, unique and cost effective wayfor companies who want to win the new challenges and maximize their CPM and BI implementations.

Application Visual Modeling

3rd Tier architecture

Master Server

Web

Client

MOLAP + ROLAP engines

What is the Toolkit Approach ?

..an integrated environment tobuild BI & CPM solutions

BOARD meets all business performance management needs, guarantying a unified access to business data,for a single version of the truth. BOARD allows building sophisticated applications for Planning, Budgeting,Forecasting, Profitability Analysis, What if scenario, Scorecarding & Dashboards, Consolidation combiningBusiness Intelligence and Corporate Performance Management on a single integrated product.

For more information: www.board.com

http://www.board.com



CubewareFounded in 1997 in Bavaria, with headquarters in Rosenheim and offices in Berlin, Darmstadt andDüsseldorf, Cubeware is one of the leading European suppliers of business intelligence solutions. Cubewaresells both a powerful and comprehensive out-of-the-box solution for analysis, planning and reporting and aconnectivity toolset for data extraction from a wide range of operational systems, including SAP. Cubeware’sproducts are platform independent, flexible and scalable, and seamlessly integrated into the worlds ofMicrosoft and SAP – as shown by the Microsoft Gold Certified Partner status and the award of SAP®Certified Integration for the Cubeware Analysis System. The products and services provided by Cubewareare targeted at the sales, finance, controlling and other specialised departments of both small-to-mediumbusinesses and major corporations. Product sales and implementation are handled either by Cubeware or byone of a growing international network of certified Cubeware business partners and resellers. OEM sales inthe form of integration of Cubeware’s BI products and components in the solutions of other BI and ERPvendors build a third, important revenue pillar.

Worldwide more than 100,000 installations of Cubeware’s analysis, reporting and controlling solutions havebeen made to date. Cubeware has been independent and self-financed from day one and employs over 50BI specialists. Cubeware has gained the confidence of hundreds of renowned customers from many lines ofindustry, including for example Abbott, ANZAG, AOK Brandenburg, Bertelsmann Stiftung, Daimler Chrysler,Danone Austria, Gabor Shoes, Kaufhof Warenhaus, Plaut Salzburg, Saeco, Viessmann and ZWILLING J. A.HENCKELS.

For more information: www.cubeware.de/eng

EPOQ GmbH

EPOQ develops and offers solutions for management and dynamic optimization of customer interactions forprocess oriented multi channel marketing. The modular product suite ready REALTIME DYNAMIC is basedon a unique method for creating dynamic scores. In outbound processes it enables dynamic customerselection, and in inbound processes it generates customer oriented product recommendations in real-time.Each customer reaction to a recommendation is fed back to the engine. This real-time closed loop optimizesthe out- and inbound activities by the continuous self learning mechanism. The proven benefits are asignificant increase of success rates, highest flexibility in campaign design, and a highly improvedexploitation of customer potentials. For enterprises that rely on high volume direct customer interactions,EPOQ offers a considerable competitive advantage.

For more information: www.epoq.de

http://www.cubeware.de/eng

http://www.epoq.de



Informatica – “The Data Integration Company”

Informatica Corporation (NASDAQ: INFA) is a leading provider of data integration software and services,solving the problem of data fragmentation across disparate systems, helping organisations gain greaterbusiness value from all their information assets. Informatica solution options include Data Quality, GridCapabilities, Structured and Unstructured Data support and Real Time data integration. Informatica's open,platform-neutral software reduces risks and costs, speeds time to results, and scales to handle dataintegration projects of any size or complexity.

Data Integration solutions for the enterprise that Informatica customers are using include but are not limitedto Data Migration, Data Consolidation, Data Synchronization, Data Warehousing, Master DataManagement, RFID Initiatives, Sarbanes-Oxley Compliance and SAP Data Migration. With Informatica,companies can gain greater business value by integrating all their information assets from across theenterprise. With a proven track record of success, Informatica helps companies and governmentorganisations of all sizes realise the full business potential of their enterprise data.

More than 2,850 companies worldwide rely on Informatica to reduce the cost and expedite the time toaddress data integration needs of any complexity and scale. In Europe Informatica has many leadingenterprises across different vertical industries in its customer base, including companies such as abbey,AUDI AG, DaimlerChrysler, Deutsche Börse, The Dutch Ministry of Defense, GlaxoSmithKline, ING DirectUK, La Poste, Mexx, Nestlé and Prudential.

For more information on Informatica please visit www.informatica.com

in-factory

is an independent consultancy company located in Winterthur, Switzerland.

Its core competency is “Enterprise Information Integration”.

For more information please visit www.in-factory.com

http://www.informatica.com

http://www.in-factory.com



Panoratio

Panoratio delivers in-memory Dynamic Data Discovery Solutions—which enable 360º customer-centricanalysis on any computer—through a patented process for rendering large, complex data sets into aPortable Data Insights™ (.pdi™). The .pdi format is transparently compressed, and ideally suited forarchiving, distribution and syndication of large amounts of data. By joining multiple PDIs together (Brick,Marketing, Web Analytics, Demographic, etc), companies can realize a complete customer view across allmajor touch-points; and can discover interrelationships and patterns that have previously remained “hidden”. All queries return in fractions of seconds, and complete analysis can occur in a fraction of the historical time.With Panoratio, there is practically no limit to the complexity of the data which can be analyzed.

Panoratio has over 50 customers in production, including Macy’s, Yahoo!, AVIS, Audi and AOL Deutschland,and is an IBM® Premier Business Partner. Please visit www.panoratio.com to learn more.

Panoratio CustomerPanoratio Customer--Centered PerformanceCentered Performance ManagementManagement

OperationalSystems

Con

vent

iona

l

…

Dat

a M

arts

SalesSales

MarketingMarketing

CustomerService

CustomerService

…

Dat

a M

arts

PerformanceManagement

Precision Segmentation

Campaign Management

Response Optimization

Churn Prevention

CLV Management…

SalesSales

MarketingMarketing

CustomerService

CustomerService

IT/Analysts

Segment A,B,C

Campaign 1,2,3

Product X,Y,Z

Region I,II,III

…

Segmentation

Campaign Management

Response Optimization

Churn Prevention

CLV Management…Fragmented Data Sources

Multiple analytic angles

Direct Sales

Complaints E-Mails

ReturnsOrders

Sales

Customer

http://www.panoratio.com



StatSoftStatSoft was founded in 1984 in the United States and is today one one of the largest global providers ofanalytic software worldwide. One of StatSoft’s flagships, the STATISTICA Data Miner, is a modern softwaretool which offers a comprehensive selection of methods for predictive analytics. It considers the proven factthat there exists no single best method for every analytical problem. But the software also provides anintuitive user interface which allows even less experienced users to create - with the help of the implementeddata miner recepies - step by step predictive models. Models easily can be deployed and implemented inother applications and run in the background. This way users get quick access to well-founded predictions.

Today the application of data mining methods is no longer limited to traditional areas like sales support, riskmanagement and customer scoring. Industry has recognized that the availability of a growing amount of datamakes complex analytics very valuable for root cause analysis and process optimization. In these applicationareas StatSoft offers specific solutions to optimize production processes. Another application area is theanalysis of unstructured text. STATISTICA Text Miner is a solution that translates unstructured text data intomeaningful, valuable clusters of decision-making "gold."

Although sometimes stated predictive analytics is not a trivial process and requires the investment of manpower and know-how. StatSoft supports its customers establishing and implementing decision supportsystems. Services will be customized to the needs of the users. Experienced consultants of StatSoft makesure that professional and appropriate solutions will be created. With a network of subsidiaries on all majormarkets on all continents StatSoft is capable to support internationally operating companies worldwide.

For more information please visit www.statsoft.com

Viscovery Software GmbHAs one of the first data mining companies in Europe, Viscovery (formerly eudaptics software gmbh) is amongthe leading vendors of predictive analytics solutions. The Viscovery suite possesses unique patentedtechnology for the explorative analysis and statistical modelling of complex data. Comprehensive workflowssupport the generation of high-performance predictive models which may be real-time integrated andupdated automatically.

For many years now, Viscovery software is being used by more than 300 customers from differentapplication areas, such as banking, insurance, telecom, industry, media, retail, as well as researchorganizations and universities. Since September 2007 Viscovery is a company of the Biomaxgroup.

For more information please visit www.viscovery.net.

http://www.statsoft.com

http://www.viscovery.net.