Capping IT Off

Capping IT Off

Opinions expressed on this blog reflect the writer’s views and not the position of the Capgemini Group

Four key challenges for Business Analytics

This article is co-written with Arild Nebb Ervik.

Regardless of how "big" the data are, success in analytics relies at least as much on organizational alignment and process as on the chosen analytical tool. Through distillation of the talk stream in the market we may identify four key challenges to address.

Key Challenge 1: Strategic Alignment Most organizations today already have some element of business analytics in place, often in the BI/data warehousing area. Unfortunately analytics are often viewed by top executives as esoteric research at best, and irrelevant fringe experiments at worst. The issue surrounds not a lack of appreciation for the usefulness of information but a lack of alignment, availability, and trust.   Recommendation: review the business goals that support the main strategies for the company, and for each major business process that underpins the goals, ask the following questions:

  • "Would we be able to govern and optimize this process more effectively if we could predict how modifications to it would affect the result?"
  • "Would we be able to adapt the process more readily to changes in the external environment if we could more accurately assess the nature and causes of those changes?"

If the answers to either of those questions is "yes", then the process (and therefore the goal it underpins, and the strategy the goal supports) would benefit from the application of analytics.

Key Challenge 2: Agility Typically in the organizations, the analysts are organized by business domains. Findings working with top-tier, information-driven companies demonstrate that domain-based organizations are not the most effective approach for analytics. Analysts often work independently and create models in ad-hoc environments based on a patchwork of extracts and sources. The results, while advanced and valid, are not easily communicated to the business users for whom they would provide the greatest value.

Recommendation: liberate the organization's analytical capabilities by pooling analysts into a Center of Excellence highly focused on "analytics-as-a-skill". Combine members of the CoE with business domain experts into teams who employ agile methods for development of analytical models, enabling your business users to gain real- or right-time insight into complex, data-demanding questions.

Key Challenge 3: Commitment Analytics software packages often come as prefabricated solution and are not particularly difficult to implement; however they can be costly, and the ROI is not immediate. By their nature, analytical models will improve in accuracy over time as the predicted results are compared with actual events hitting the warehouse. But this is a complex endeavor that requires dedication to the solution during an extended tuning period. Here is where many deployments fail. Business users do not immediately see the promised results and lose interest, and executives lose trust in the solution and refuse to rely on what the models tell them.

Recommendation: by addressing key challenge 1, stakeholders in analytics will naturally be identified (the process and goal owners). These business owners must take responsibility for establishing the productive analytics environment described in key challenge 2. Realistic timelines that allow the models to take form should be set based on industry standard and best practice.

Key Challenge 4: Information Maturity The world's best hammer is useless without nails, and so it goes for analytical tools. For an analytics solution to succeed, the "nails" need to be plentiful and not consistently bent or misshapen. Implementations often fail because of the lack or low quality of underlying transactional data. Either data are not available, data sources are too complex, or data are poorly mastered. Even bleeding-edge, sentiment- and context-analysis tools require some level of trust in the data, and for any analytical model the rule is consistent: the more trustworthy the data the more trustworthy the result.

Recommendation: perform a maturity assessment on the company's information architecture. Identify data sources based on a mapping to analytical requirements; measure the quality of both operational information (transactional data) and aggregated information in the warehouse; and review the existing integration infrastructure's ability to support new sources and data conduits.

Conclusion Analytics cannot be adopted for its own sake or as a layer on an existing BI infrastructure. It requires a coherent, dedicated approach and a decent level of information maturity.

About the author


Leave a comment

Your email address will not be published. Required fields are marked *.