Ten top reasons why BI implementations fail

Business Intelligence (BI), when implemented correctly, allows businesses to reap significant benefits and gain valuable insight into their organisations based on actionable data. But there is much more to BI than simply deploying technology. Organisations need a comprehensive, strategic approach to designing, implementing, managing, tracking and supporting BI initiatives, says Goran Dragosavac, practice lead for analytics at SAS Institute South Africa

Lacking that framework, the organisation could end up with an expensive patchwork of good intentions but no meaningful enterprise-wide intelligence. Here are some of the common pitfalls:

1. Poor Requirements Gathering

It is very important that BI vendors guide organisations through the exact requirements to meet their needs. This helps to avoid spending copious man hours extracting and collating data that might not be relevant to the design of their BI architectures. Effective BI is all about the underlying data and what it is going to analyse. The data must be accurate, consistent and trustworthy for the purposes that it will be used for. Yet at this extremely important level, few companies have strong requirement gathering practices.

2. Scope Creep

In the design and requirement phase of most BI implementations organisations often understand that BI can solve many problems and provide insight into key areas that drive the business. However, incorporating too many checks and balances could turn it into a mammoth project and in the end might not provide the deliverables that the users expected in the first place. BI vendors know that user expectations have to be managed and understood by those who are tasked with identifying the deliverables of the project.

3. Poor Design & Implementation

When implementing BI, a poor development effort can be corrected, but if a system is poorly designed from the bottom up, it would require a complete make-over as data volumes grow and new analysis functions are introduced. Companies must explore their intentions when implementing a solution and look at the bigger picture, not just its immediate function. In the design phase, it is also paramount that organisations be mindful of the metadata strategy – where the data originated, what it is and how it will be used.

4. Large Projects vs. Small Steps

When implementing BI, the golden rule is to think big, but to start with small meaningful implementations that have a direct measurable impact on the organisation’s bottom line. This allows users to get accustomed to the system and provide feedback on what could be improved or discarded.  Many companies that embark on large BI initiatives that impact all aspects of the business spend large amounts of money on a big project that will take years to complete, and which eventually is abandoned only a few years later. This affects top management buy-in for future BI initiatives, especially in today's tumultuous economic environment where IT budgets are tighter than ever.

5. Impulsive BI Application Selection Process

Whilst many companies realise that BI can provide them a distinct advantage over their competition, it is important to align the needs of the business with the functionality that BI can provide. Many BI vendors also claim that their platforms can perform varied functionality, so it is extremely important that organisations go through a proof of concept stage to ascertain if the software is right for their business. In this way, organisations are far more likely to achieve success and meet user expectations. Often, organisations choose a BI platform that does provide the functionality they require, but is cumbersome to use because of the extensive development that was required to make it work.

6. Inadequate Functional Testing

Functional testing is a fundamental part of any implementation to ensure that the application fits all the requirements before it gets rolled out to a broader set of users. However, in many BI implementations, this step is ignored and still doesn't provide users with what they need, which decreases user adoption and can ultimately lead to the failure of the project.

7. Performance Problems

With new implementations, user expectations are generally very high, yet many organisations do not have the infrastructures in place to ensure that the new BI system can perform at adequate speed whilst accommodating the amount of users in the organisation. If the system is slow or unstable, users become frustrated and will revert to the ‘old way’ of doing their daily jobs. It is therefore important that the application is thoroughly tested before it is introduced to the users.

8. Inadequate User Involvement and Training

Effective use of information to drive decisions requires that the organisation have the necessary skills, culture, infrastructure and processes in place. In other words, there must be an environment of readiness. Employee willingness to change roles and responsibilities is important to the BI culture because use of information to drive decisions will require workers to gain skills necessary to understand the results of data analysis and apply those to the business.

But many organisations still lack some of these essentials necessary to make the move to an enterprise BI strategy. While employees are open to changing roles and responsibilities, the real barriers are data and information availability and access, which seem to be compounded by a siloed approach to information and lack of sharing across the organisation. A support centre for BI training and assistance is often the hallmark of organisations that have made a full commitment to intelligence-driven decision making.

9. Non-conformance on Data Cleansing

Most organisations capture data in a wide variety of sources and formats, including enterprise resource planning (ERP), legacy systems, relational database management systems (RDBMSs), flat files, Web logs, and so on.

This diversity creates challenges to locating, identifying and selecting the right data. Rarely are there discussions about data that are not accompanied by concerns about unwieldy amounts of data, lack of consistency across sources and differing data definitions.

Effective BI is all about the underlying data. It must be accurate, consistent and trustworthy. The fundamental components of effective BI – data quality, integration and consistency – are among the most neglected.

10. Inadequate Sponsorship

If an organisation has made the decision to implement a BI strategy and rely on business information to drive its decisions, the success of that strategy will depend in part on the organisation’s level of commitment – and the degree to which management actually uses analytics-driven intelligence to support decisions. Tangible support from executives – commitment evidenced by action – sets the tone for the rest of the organisation.

At the end of the day any BI implementation requires careful consideration before an organisation rushes into huge capital spend and a project that may very well be abandoned before completion. By addressing the top reasons why these projects fail companies can set themselves up for future success and can then drive the benefits of sound BI throughout the organisation.