12
Final Project - BI/DW Solution Presented to Dr. Turel ISDS 556 Prepared by Dhirendra Singh Jiahe Li Ning He 1

ISDS 556 - Final Project

Embed Size (px)

Citation preview

Page 1: ISDS 556 - Final Project

Final Project - BI/DW Solution

Presented to

Dr. Turel

ISDS 556

Prepared byDhirendra Singh

Jiahe LiNing He

Szu Tung ChenSarun Sangarungroj

Fall 2014

1

Page 2: ISDS 556 - Final Project

Table of Contents

I. Overview.................................................................................................................................................. 3

II. Description of the Organization and Need for BI...................................................................3

III. Description of the Functionality and Features of the Required BI System ...............3

IV. Data Model.............................................................................................................................................. 3

V. Data Analysis......................................................................................................................................... 5

VI. Extraction, Transformation, and Loading Issues...................................................................5

VII. The Recommended Architecture..................................................................................................6

VIII. Preliminary Budget Plan...................................................................................................................7

IX. Conclusion............................................................................................................................................... 8

2

Page 3: ISDS 556 - Final Project

I. Overview

Target Enterprises (Pty) Ltd. is a mail order company that sells a variety of consumer items. The project is about measuring Order Processing process. It is needed because the company need to analyze several aspects of the Ordering Process, such as how many orders are coming every day, how many orders are received by mail, fax, or phone, which item is in demand, and which state/county/city is placing most of the orders. The BI system is needed because it would help them know what is going on within the business.

II. Description of the Organization and Need for BI

Target Enterprises (Pty) Ltd. is a mail order company that sells a variety of consumer items. We assume that currently, the company does not have any BI/reporting system. However, it is not possible to analyze operational data. The proposed system would be a new BI system for Target Enterprises (Pty) Ltd., and it’s ready for the new BI systems. Here is the high- level enterprise data warehouse bus matrix for Target Enterprises (Pty) Ltd.

Dimensions

III. Description of the Functionality and Features of the Required BI system

We are focusing on Order Processing. Because we assume that this process is the first process in the value chain and also the most important process for the company. In addition, the KPI of the BI system would be number of orders, quantity of items and cost of items.

Many types of transactional as well as summarized reports can be generated out of the system.  Both summarized reports and transaction level reports would be used by upper management while the staff would be given access on transaction level reports. The reports could be run at any time. The BI application would fetch the data for presentation/analysis.

IV. Data Model

We have developed a star schema for our data model, in which there is one fact table and 7 dimension tables. In the Customer and product dimension table all the attributes are being treated as Type-2, while in other dimension tables, all the attributes are being treated as Type-1. The attributes "Order Number" and "Order Line Number" are being treated as degenerate dimensions in the fact table.

3

Page 4: ISDS 556 - Final Project

Note: The Date Dimension is a role playing dimension.

4

Page 5: ISDS 556 - Final Project

5

Page 6: ISDS 556 - Final Project

V. Data Analysis

In our data model, there is one record per order line per order, and it means that if there were two items, item A and item B, on an order, there would be two records in the fact table. The data would be coming from operational system. The data source is appropriate and the data doesn't need much cleansing. With this data model many types of reports can be generated as given below.

1. Total number of orders in a particular day/month/year for any specific product and/or overall.

2. What products are the best selling products by month/customer?3. What payment type did customers most commonly use?4. Which clerk handled the most orders in different months?5. What order types, such as fax, phone and mail was mostly used by customers by

day/months/year?6. What was the total sale by the store by day/month/year?7. What was the total sale of a particular product by day/month/year?8. Who are the best customers in terms of the amount of purchase by

day/month/year??9. Customers from which state placed the most number of orders by

day/month/year??

Like that many other reports can be generated using this data model.

VI. Extraction, Transformation, and Loading Issues

Data extraction might be an issue even though the data model is quite simple. The extraction could take a lot of time because the amount of data from both internal and external data. Applying validation check to the user interface (where users input the data) could at least reduce the formatting problem in the data, thus faster extraction.

Since all the data come from an operational system, it should be mostly consistent. We might want to have a validity check on customer phone number because the formatting could be different for each customer. Some might put parenthesis around area code or dash between chunks of numbers. E.g. (xxx)xxx-xxxx while some might put only just numbers. Aggregations are also needed because we need to show many summary stats like the most sold item, the best customers by amount of purchases. Also at the end, all data will be sorted.

The best timing to extract and/or load data depends on the requirements of each organization that want to utilize the data. In general, we do not need real-time loading since the model is built to provide summaries by day/month/year, so daily would be the best choice.

6

Page 7: ISDS 556 - Final Project

VII. The Recommended Architecture  

The data source contains internal data and external data. The internal data includes the data from source systems and operational systems, such as OLTP system and ER system from different stores. And the external data includes the data from Target Enterprises (Pty) Ltd.’s product suppliers. Next, the ETL system will extract, transform, and load all the data into data staging area. ETL system will integrate and process all the data from multiple systems in different stores. For example, looking up the inventory status in one store while updating the inventory status in other stores, sorting the new sales file while removing the duplicates on another file, checking order status for specific order, or loading the historical sales data while creating the new sales data.

Metadata, summary data, and raw data describe stage in presentation area. Metadata summarizes basic information about data, which can make finding and working with particular instances of data easier. For example, retrieving the data of monthly sales of each store and summarizing sales quantity of each product.

In the end, the end users can use BI applications, such as OLAP analysis, reporting, and data mining skills for getting the sales and inventory reports from each store, finding potential patterns of sales, predicting the order cycle of each supplier, summarizing the average time the Target Enterprises (Pty) Ltd. took for picking, packing and delivery of the packed items to a shipping carrier.

Therefore, Target Enterprises (Pty) Ltd.’s could use this application architecture model for improving its order processing procedure. First, it could analyze multidimensional data interactively from multiple perspectives by OLAP analysis. Second, it could get detailed sales summary reports and specific reports on each order process by reporting BI application. Third, it could use data mining skills to develop other data analysis. For

7

Page 8: ISDS 556 - Final Project

example, classifying frequent buyers, predicting future sales, finding association rules on its products, and creating multidimensional visualization charts for upper managers

In the end, the BI applications could benefit Target Enterprises (Pty) Ltd by decreasing order processing time, increasing sales for its company, creating value on its products, and further expanding its reputation.

VIII. Preliminary Budget Plan

Since Target Enterprise is a small-sized mail company with twenty employees. It is not necessary to implement a large expensive Business Intelligence system. Instead, we recommend Birst, a cloud based BI leader.  Birst provides automated routines for standard data integration, and ETL for more complex needs. The package costs about $30,000 per year.

In order to build a DW/BI system, the company also needs to form a BI group. Because this BI project is for one part of daily business in a small company, we do not need a large BI group. There will be four people in this group: manager, business analyst, ETL developer, and BI developer. The manager overlooks a whole project. The business analyst will do research for find out business requirement.  He/she could be selected from current marketing group.

The ETL developer is responsible for extract, transform and load data from related operational department to the BI system. He/she will also work for other backroom tasks. The BI developers are responsible for overall front room in the system. The recruitment for three new employees could cause a salary increase of about $35,000 per year.

Although implementing of the new DW/BI system will lead to extra costs of $380,000 per year, it helps to save money in other areas. First, senior manager could make decision more efficiently based on consolidated reports generated by the new system. For example, by reading those reports, managers could get knowledge of each market more accurately.

Therefore, they could focus market campaign in certain areas instead of wasting money in those useless activities. Second, the system help the company to get insight into customer behavior, it is unnecessary to hire an external consultant to do this kind of research any more. Moreover, since BI generates key business reports when and where you need them, it saves money to hire someone else to generate the reports. The total savings are expected to be $450,000 per year.

After introducing DW/BI, the company could have a yearly saving of about $70,000. Although it is a small number, it could not be neglected. As to the benefits it will bring to the business, the company is expected to generate higher revenue in the following years.

8

Page 9: ISDS 556 - Final Project

IX. Conclusion

As BI has become an increasingly popular and accepted technology for improving business performance, even a small company like Target Enterprises can greatly benefit from implementing BI system. The new BI system can help the company get faster answers to business questions, get key business metrics reports (transactional and summarized) and also get insight into customer behavior. It is estimated that after introducing the BI system, the company could achieve a significant amount of yearly saving.

Successful deployment and operation of the BI system requires thoughtful planning and coordination prior to the completion of development efforts. In our report we talked about how the features and functionalities of the BI model and it can help meet the organization's operational and analytical needs. Also we designed the data model, created the application architecture and talked about how we can address crucial issues related to extraction, transformation and loading process. In the end we provided recommendation on application tools and developed a budget plan.

Our next steps would be system deployment and maintenance.  Throughout the deployment process the system will be tested extensively. In addition, documentation and user training will be required. After the system is deployed we need to provide maintenance and support, including front room and back room, to ensure the system runs properly. In the future if the business expands, we should also consider expanding the current system.

9