Gartner Businness Analytics

Table 1. Agility for Business Analytics

Figure 1. Analytic Process Agility

Figure 2. Conceptual Approach for BI and Analytics Capability

How to Be Agile With Business Analytics27 March 2015 ID:G00270535

Analyst(s): Alan D. Duncan, Alexander Linden, Thomas W. Oestreich

VIEW SUMMARY

As the demands of IT service delivery shift toward bimodal capabilities, analytics and business intelligence leaders must introduce increased agility into the business analytics delivery process, in order to increase the agility of the business decision-making process and maximize business value.

OverviewKey Challenges

While disruption within the business intelligence (BI) and analytics market makes this area ideal for early adoption of bimodal IT practices, businesses are struggling to achieve the desired levels of responsiveness and flexibility.

"Analytic agility" has three complementary facets, with each facet needing to be considered differently; and to achieve true agility within business analytics, all facets need to be addressed.

The term "agile" is used within the IT domain to describe a specific category of software development methodologies, whereas agility within analytics practice, refers to a general capability to be responsive, flexible and deliver fast time to insight. This leads to confusion.

RecommendationsAnalytics and BI leaders should:

Take steps to increase the agility of BI and analytics processes.

Build BI and analytics teams with the process, cultural and architectural capabilities for agility, and embrace business participation in analytics processes.

Invest in BI and analytic environments for experimentation and evidence-based data discovery. Then follow up with resilient operationalized analytics solutions where repeatable value is identified.

TABLE OF CONTENTS

CONTENTS

IntroductionAnalysis

Improve Agility of Analytic ProcessesInformation PortalAnalytics WorkbenchData Science Laboratory

Build BI and Analytics Teams With a Culture of Analytic AgilityAnalytic Team ConfigurationTeam LeadershipTeam Makeup and SkillsData Science Labs Empower Data Scientists

Invest in BI and Analytic Environments for Experimentation and Evidence-Based DiscoveryData Lakes Can Be Side-Line BarriersOpen-Source ToolsSmart Data Discovery ToolsCloud-Based Analytics Services

TABLES

FIGURES

STRATEGIC PLANNING ASSUMPTIONS

Through 2017, 90% of the information assets from big data analytic efforts will be siloed and unleveragable

across multiple business processes.1

By 2017, 75% of IT organizations will have a bimodal

capability. Half will make a mess.2

By 2017, most business users and analysts in organizations will have access to self-service tools to

prepare data for analysis.3

ACRONYM KEY AND GLOSSARY TERMS

API application programming interface

BI business intelligence

CRAN Comprehensive R Archive Network

KNIME Konstanz Information Miner

KPI key performance indicator

REST representational state transfer

EVIDENCE

1 "Predicts 2014: Big Data."

2 "Predicts 2015: Bimodal IT Is a Critical Capability for CIOs."

3 "Predicts 2015: Power Shift in Business Intelligence and Analytics Will Fuel Disruption."

4 For further information visit the website, Manifesto for Agile Software Development.

5 "Gartner Digital Business Baseline Survey 2014."

6 Based on analyst inquiry interactions with Gartner clients.

Página 1 de 7How to Be Agile With Business Analytics

16/08/2015http://www.gartner.com/technology/reprints.do?id=1-2H7KPS5&ct=150604&st=sb

IntroductionThe digitalization of 21st-century business — and the combination of information, social, mobile and cloud computing — requires increased focus on responding and adapting to change. This requires the IT function to be much quicker and more adaptive in supporting these changing business dynamics, with business intelligence (BI) and analytics at the forefront of this shift. Business analytics is a stimulus for innovation, with evidence-based decision-making paramount — in both strategic and operational business execution. Meanwhile, traditional BI and analytic models are being disrupted, as the balance of power shifts from IT to the business.

Within this context, analytic agility is the ability for business intelligence and analytics to be fast, responsive, flexible and adaptable. Time to value is paramount for fact-based evidence that supports the business decision-making process. Improved analytic agility supports more nimble business models, even in circumstances where the business decision-making process is not formally defined.

Analytic agility therefore means adopting a bimodal IT approach within the analytics team (see "Predicts 2015: Bimodal IT Is a Critical Capability for CIOs"). As outlined in Table 1, the development of analytic agility needs to be developed and embedded across three complementary analytics capabilities — the technology and architecture, the analytic process and the skills of the analytics team.

Table 1. Agility for Business Analytics

Analytic Agility Capabilities Features Enabling Agility Features Inhibiting Agility

Technology and Architecture Reusable

Adaptive

Repeatable

Scalable

Iterative

Open

Modular

Sandbox

Single Purpose

Specified in Detail

Proprietary

Rigid

Monolithic Data Warehouses

Analytic Process Flexible

Bidirectional

Continuous Learning

Use Case Specific

Process-Bound

Constrained

Controlled

Analytic Team Responsive

Adaptable

Courageous

Democratized

Rigid

Inflexible

Risk Averse

Centralized

Source: Gartner (March 2015)

Note: In some cases, increased analytic agility may entail adopting agile principles (as defined by the Agile Manifesto) and associated methodologies, for example, Scrum or Dynamic Systems Development

Method.4 See "Best Practices in Transitioning to Agile: Picking a Methodology" and "Best Practices for Implementing Enterprise Agile Principles." For the purposes of this research, adoption of agile methods is not considered a prerequisite for increased analytic agility more generally.

AnalysisImprove Agility of Analytic Processes Agile analytic processes benefit from leveraging the key principles of agile software development, adapted to transforming data into insight that supports and improves business decisions. The process steps of analytic workflows are common across different use cases of analytics, although the way they are applied can vary for different functions.

The key analytic process steps are:

Identify and understand the business objectives, issues and key questions.

Acquire data, including identifying and getting access to required data sources.

Store and transform the data into data models that can be analyzed.

Process the data and apply analytic methods, algorithms and rules.

Investigate the developed insight and present as data visualizations.

Review and interpret the insight gained.

Execute the required actions; align with the business objectives.

Figure 1. Analytic Process Agility




(See also "Aspire: A Framework for Analytic Business Processes" and "How to Deliver Data Discovery Projects.")

Gartner recommends setting up a three-tier approach for a BI and analytics capability in order to serve the different analytic use cases:

1. Information portal2. Analytics workbench3. Data science laboratory

Figure 2. Conceptual Approach for BI and Analytics Capability

BI = business intelligence




Analytic agility requires that the tiers do not operate in isolation. Interoperability and compatibility between tiers is important. BI content and analytic artifacts should be promotable and easily interchangeable between the tiers, making them reusable.

(See also "How to Architect the BI and Analytics Platform.")

Information Portal This is usually built upon a central enterprise BI platform and data warehouse, managed by the IT function. Business users have access to prebuilt reports and dashboards, based on key trusted metrics and performance measures. Analytic output is the result of a formal development process, with delivery times dependent on complexity and workload. The information can be trusted and is used across the organization, but has less flexibility and reduced interactivity.

Within the information portal, increased agility requires an adaptive and responsive technology infrastructure. Quickly responding to new requirements is based upon:

Master data management processes that ensure consistency of data.

Reusable key performance indicator (KPI) and metrics frameworks.

Disciplined information life cycle management, avoiding oversized and over-complex data warehouses.

Constant review of existing analytic content and artifacts to avoid redundant reports and dashboards.

Self-service BI components for business users to perform their own analysis.

Analytics Workbench This is the workspace used to investigate trends and interactively visualize data. It is the tier where data discovery and ad hoc query tools are used to explore information with access to a broad range of data sources. These are typically deployed as decentralized solutions with limited involvement from IT. The idea here is to give business analysts a high degree of freedom in working with data from various sources to derive both descriptive and diagnostic analytics.

The analytics workbench largely supports ad hoc requirements. "Where is the problem?" and "Why did this happen?" are typical questions addressed. Agility in the analytics workbench is based upon:

Business analysts performing all analytic process steps on their own.

Enabling business analysts to easily mashup the data sources and perform data preparation.

Supporting storytelling, sharing and collaborating on analytics with other users.

The capability to promote content developed by the business analyst to the information portal.

Data Science Laboratory This is the workspace where advanced analytics takes place and is the ideal incubator for big data initiatives. It is a flexible environment where experimentation, innovation, creativity, trial-and-error, and "fail fast, fail often, learn fast" are all encouraged. A broad set of technical capabilities is provided by specialized tools with minimal IT integration, delivering the ability to respond to unforeseen questions. Users are skilled and experienced, often even more so than the technical resources in IT.

Agility in the data science lab should be achieved by:

Focusing on experimentation, based on hypothesis derived from ideas — not following the formal standard analytic processes.

Performing short analytics processes and multiple iterations.

Running experiments with a fail-fast attitude — avoid binding resource too long on a specific project.

Enabling data science teams to use the best-suited technologies, irrespective of corporate standards.

Build BI and Analytics Teams With a Culture of Analytic Agility Overall, success of evidence-based decision making is largely predicated upon the ability of the analytics team to engage and deliver the analytics services as well as support an evidence-based culture.

Analytic Team Configuration Gartner research shows that the biggest challenges organizations face when implementing a digital strategy are integrating adapting digital capabilities into business processes (58% of respondents) and

faster implementation (48%).5 While fixed, linear delivery and task-oriented planning are suitable where robustness, operational repeatability and auditability are required, executives need evidence to inform and guide their thinking — and they need it quickly. Yet Gartner clients are still reporting lead

times of six weeks or more to develop and deliver a new business report.6 This shows that traditional "IT factory" models based on strict task-based assignment of jobs and tasks using hierarchical working practices are not succeeding in achieving the levels of agility that digital business requires. Challenges include:

Many analytical scenarios that are oriented toward specific one-off management decisions, often in response to a change or market conditions, or as a new potential opportunity arises.

Timescales, since there may be little time to work out a detailed task-based plan, requiring an ad hoc approach based on collaborative working, flexibility and problem solving.



It being likely that the team will be working across multiple analytic models, with conflicting priorities. It is challenging to evaluate the relative priorities of each of these initiatives; setting one project against another can create organizational conflict and be counter-productive to the benefit of the organization as a whole.

To overcome these challenges requires an analytics team structure that enables and encourages the team to adopt different roles, dependent upon the nature of each specific project. Factors may include:

Depending on the organizational culture and dynamics, the analytic team may either be fully centralized or else established in a more "virtual" manner with collaborative and cooperative communities of practice. (See "Organizational Principles for Placing Advanced Analytics and Data Science Teams.")

That once in place, the analytics team may move to become more self-organizing and make its own decisions about who will fulfill which role in order to achieve the required project outcomes.

Resourcing levels that may need to be variable to flex with changing levels of demand — a core analytics team augmented by point resources from trusted third parties may be a suitable approach to bringing skilled resources into the team quickly when required.

The analytics team that may also co-opt other internal resources when required, and work in partnership with the business community, delivering on the basis of shared values and shared outcomes.

This flexibility of approach that also enables the team to work on multiple projects in parallel, shifting work assignment and priority as each item becomes less or more urgent, and less or more important. (See "Maverick* Research: Fire Two-Thirds of Your IT Organization.")

Team Leadership Self-organizing analytics teams are not totally autonomous. The analytics center of excellence manager must create an environment where the team has the delegated authority and freedom to operate effectively and deliver value.

The manager will set an overall scope and be responsible for any boundary conditions and constraints. They are also accountable for the team's performance, so monitoring achievement of goals and redirecting the team's efforts when required are also crucial.

Like other leaders of self-organizing teams, analytics team leaders will display characteristics including:

Relating: Being socially and politically aware; building team trust and caring about team members.

Scouting: Seeking information from other stakeholders; understanding team behaviors and systemic investigation of problems.

Persuading: Engaging and obtaining support; encouraging the team.

Empowering: Delegating and supporting the team; flexible decision making; coaching.

(See also V.U. Druskat and J.V. Wheeler, "How to Lead a Self-Managing Team," MIT Sloan Management Review, Summer 2004.)

Team Makeup and Skills Members of the analytics team need to have excellent technical skills and be proficient with multiple tools, packages and algorithms. They will also have a strong affinity for data and quantitative analysis methods.

However, the success of the team in achieving analytic agility is not solely vested in the team's technical abilities. Human engagement and an ability to communicate with stakeholders is the factor that will determine overall level of success. Characteristics of analytically agile team members will therefore include:

Critical Thinking: The analytics team makeup should consist of staff who are problem solvers and critical thinkers. They need to be flexible and adaptable, creative and innovative, and have a capacity to embrace change, learn and adapt.

Communication: The ability to engage, communicate and tell data stories is paramount. As an advocate for the exploitation of information assets for evidence-based decision making, an analyst or data scientist will be able to relate the insight gained from data into a meaningful business narrative, as well as convey the business impact in a way that is meaningful to the business community — whose role it is to then take the desired action.

Outcome Focus: A focus on delivery is also crucial — a "can-do" attitude is vital. Behavior, mindset and a culture of responsiveness mean that members of the team will be outcome-oriented and value-driven. This means real and thorough understanding of the business model and key processes, paired with an ability to consult, curate, facilitate and coach.

Comfortable With Bimodal Work Practices: Analytic agility requires that team members be collaborative and open-minded, responsive and adaptable. They will have the ability to work with speed when needed, and with rigor when necessary. They will be comfortable with "good enough" since perfectionism inhibits responsiveness and creativity. However, they will also be mindful of the need for traceability and auditability when appropriate, coupled with the ability to deliver robustly engineered analytics solutions when required to do so.

Rapid Prototyping: Rapid development of working analytics prototypes using real data enables enhanced experimentation and "fail fast" approach to innovation insight. Data pipelining is a primary way of accelerating construction of such prototypes. In this approach, building-block components present predefined data sources, data preparation functions, data discovery tools, visualization and even embeddable advanced analytics algorithms. These can be quickly arranged to meet the functional requirements of the analytic objective. Applying data pipelining, agile teams



will consciously trade overall precision and perfection against transparency, ability to communicate to different stakeholders, as well as the reuse and significant acceleration to project delivery.

See also "Bimodal IT: How to Be Digitally Agile Without Making a Mess."

Data Science Labs Empower Data Scientists In organizations where data scientists are employed, these roles are typically scattered throughout the enterprise and operate in a localized and uncoordinated manner. In contrast, the notion of a data science laboratory proposes to create better empowerment of data scientists by establishing a "critical mass" of capability that can drive agility through:

Sharing of data engineers (for data sourcing, and data preparation).

Discussion of common goals (for example, data governance).

Acknowledgment of common principles and technologies.

Vetting new projects.

(See also "Extend Your Portfolio of Analytics Capabilities.")

Invest in BI and Analytic Environments for Experimentation and Evidence-Based Discovery There are several choices with regards to analytics architectures, infrastructures and tools that will further drive and support agility in the analytics center of excellence.

Data Lakes Can Be Side-Line Barriers Data lakes are free-form, nonschematic data repositories, where the contextual definition of data is inferred programmatically at the point of need from the content of the dataset. This is done using a machine-executed solution — a "schema on read" approach. (This is in contrast with scenarios where data is already presented in a readily structured form by writing the data to predefined underlying logical and physical data models or schemas. This "schema on write" approach has been a predicate condition of traditional data warehousing and business intelligence architectures.)

This sort of late-binding gives tremendous flexibility and makes data lakes a must-have for any agile analytics or data science team (see "Hype Cycle for Advanced Analytics and Data Science, 2014"). Data lakes facilitate ad hoc and cross-functional inductive analytics on a level of detail that would be virtually impossible via data warehouses. Especially, inductive analytics requires very granular, cross-functional data and access to novel datasets, which data warehouses most often will not provide.

However, caution should still be applied, as such capabilities are still largely immature and only weakly supported by vendors. In particular, the security concerns of largely ungoverned centralized data repositories must be approached with caution. (See "The Data Lake Fallacy: All Water and Little Substance.")

Open-Source Tools Extremely agile teams currently have a strong emphasis on open-source tools such as Apache Hadoop, Apache Spark, R, Python, KNIME, and RapidMiner. These can offer significantly lower cost outlays for core software and also are at the forefront of new advances in techniques such as deep learning. New open-source function libraries are made available much more quickly (for example, the Comprehensive R Archive Network [CRAN], scikit-learn) than the well-policed corporate packaged applications.

The inquiry feedback Gartner receives from our advanced analytics clients is consistent:

"open-source R or Python allow me to do the same things or even more — in a much more flexible environment, while saving often millions of dollars in annual software licenses, which I can spend on expanding my agile team"

Smart Data Discovery Tools Smart data discovery tools facilitate the discovery of hidden patterns in large, complex datasets, without building models or writing algorithms or queries (for example, Ayasdi, BeyondCore, Emcien, IBM Watson Analytics, SAS Visual Analytics).

With smart data discovery, less experienced data scientists can still benefit from advanced analytics to highlight and visualize important findings, correlations, clusters, links or trends in data that are relevant to the user. User interaction and exploration is provided via interactive visualizations, search and natural-language query technologies.

Analytic teams should give consideration to these tools because they can free up data scientists in two specific ways:

1. Relief from tedious programming through advanced analytics platforms or using the open-source stack.

2. Allowing other "citizen" data scientists to take on some of their work, for example, initial exploration of data.

(See also "Predicts 2015: A Step Change in the Industrialization of Advanced Analytics.")

Cloud-Based Analytics Services The cloud has much to offer in the area of analytics, especially advanced analytics. The evolving orchestrated analytics ecosystems will provide significant opportunities for much greater agility (with, for example, Microsoft Azure Machine Learning cloud at the forefront):

Seamless upscaling the required processing power.



In-the-cloud provided with the ability to trigger/monitor/control time-consuming events overnight and over the weekend — and from wherever you are.

In-the-cloud provides scattered teams across the globe easy access to the same data science experiments.

The deployment features of in-the-cloud only require the publishing of a representational state transfer (REST) API, together with the consumption of the same REST API on the receiving side of the analytics models.

© 2015 Gartner, Inc. and/or its affiliates. All rights reserved. Gartner is a registered trademark of Gartner, Inc. or its affiliates. This publication may not be reproduced or distributed in any form without Gartner’s prior written permission. If you are authorized to access this publication, your use of it is subject to the Usage Guidelines for Gartner Services posted on gartner.com. The information contained in this publication has been obtained from sources believed to be reliable. Gartner disclaims all warranties as to the accuracy, completeness or adequacy of such information and shall have no liability for errors, omissions or inadequacies in such information. This publication consists of the opinions of Gartner’s research organization and should not be construed as statements of fact. The opinions expressed herein are subject to change without notice. Although Gartner research may include a discussion of related legal issues, Gartner does not provide legal advice or services and its research should not be construed or used as such. Gartner is a public company, and its shareholders may include firms and funds that have financial interests in entities covered in Gartner research. Gartner’s Board of Directors may include senior managers of these firms or funds. Gartner research is produced independently by its research organization without input or influence from these firms, funds or their managers. For further information on the independence and integrity of Gartner research, see “Guiding Principles on Independence and Objectivity.”

About Gartner | Careers | Newsroom | Policies | Site Index | IT Glossary | Contact Gartner



Documents

Gartner Businness Analytics