Implementation of BI: Useful guidelines for effective execution

Proper implementation of a business intelligence (BI) project results in numerous advantages. Here are a few tips for successful execution.

Implementing a BI project involves considerable investments in terms of time and money. Hence, it is important to ensure that its flawless execution. We highlight some key considerations to be borne in mind while undertaking a BI implementation.

1) Set only the right expectations: It is unrealistic for the business to expect that BI implementation will work overnight wonders in terms of data management and automate existing reporting processes. Instead, a BI project should be initiated with core business objectives like increasing customer base and satisfaction, expanding existing markets, and growing revenues, among others. Before initiating any project module, it is important to consider factors such as its relevance, accuracy, consistency, and timeliness.  

2) Ensure executive sponsorship: It would be easier to get executive sponsorship if the implementation of BI is based on business tenets. The requirements of each department in the organization should be placed before the steering committee (which involves department heads) and regularly review the execution process. The IT department should only be responsible for implementing BI’s technological aspects.  

3) Clean the data: Data is the most critical raw material to make a BI solution work. It is therefore, important to process and evaluate usable information from data. However, one of the biggest challenges in the implementation of BI is deciding who is responsible for cleansing of data – the IT team or system owners? Hence, it is important to clearly demarcate the duties of the IT team and system owners before data starts accumulating.

4) Data collection: Assuming that data can be easily acquired from the online transaction processing (OLTP) systems could result in delays during implementation of BI. While designing the extract, transform, and load (ETL) procedures, look carefully at every field, check how the data is captured, and determine update mechanisms for the field. Every organization has ‘Excel jockeys’, people sitting on secret piles of data. Their data formats need to be acquired and standardized, as well as perform a onetime porting of many of these Excel sheets into tables. The departmental data mart is more evolved, and gets its inputs from the main enterprise applications. If a concerted effort is not undertaken to identify such entities, you may have to feed the data mart from your BI stack, rather than the other way round. You should also check if the BI project is capable of fulfilling the data mart’s function, which may help reduce the need to maintain one more application.

5) ETL is the nucleus: Your ETL tool should be flexible. If possible, create a detailed internal training process or get external help. You may also create user manuals and a support mechanism during the rollout’s first few weeks. These people should be able to check the work of your software vendor.

6) Project documentation: While undertaking BI implementation, it is important to ensure that project documentation, especially the ETL procedures, are clearly documented with proper version management. It is of immense utility to create a knowledge base of all reported issues and the used solutions.

7) Customize, with care: Customization is in demand, but are you prepared for the corresponding long term support? It is better to generalize rather than customize. For example, technology allows you to serve cubes from rolled up dimensions, from which users can create their own reports. Evangelize such approaches, rather than point solutions.

8) Get the team on the same page: While collecting data for BI implementation, it is important to ensure that it is on the same page with reference to the terminologies. Different teams in the same organization could have varying definitions for the same term. For example, even a common term such as ‘margin’ may elicit different numbers from your procurement and finance teams. Since the premise of enterprise BI stack is to deliver a single version of truth, it is essential to have a single version of the terminology. The project has to be communicated as a means of making work tractable.

9) Choose the vendor judiciously: The leading IT service providers have been consolidating in the past few months — presumably to offer their customers with a one-stop solution. However, be prepared to deploy different presentation layers as part of your BI implementation. While one tool may have excellent reporting features, it may not have great dash board features. You may also be unaware of solutions that shorten the BI implementation time. For example, during Shoppers Stop’s BI implementation, it purchased a retail data model from a boutique vendor. The data model basically had a blueprint of fields for the data warehouse, storage formats, and a list of metrics used by international metrics, which immensely helped the organization. Hence, evaluate every solution thoroughly — if possible, do a proof of concept (PoC) on your own data to test parameters like performance.

Irrespective of the SLA’s clauses, your BI project’s success depends on resources deployed by the vendor. Hence, know exactly what you want, and don’t expect your vendor to have all the answers. Depend on your vendor only for technical inputs, as domain knowledge is your competence. Also, keep a track on the solution updates available in the market.

About the tip: This content piece has been compiled from Ranjit Satyanath’s (Head - Technology at Shoppers Stop & Crossword) presentation on “BI - How NOT to implement it” at a recent event. It has been collated by Sharon D’Souza.

Read more on Business intelligence software