Response Design Corporation: Kathryn's Blog

As little as a few years ago I heard executives tout the great numbers being presented after data mining. The result of the mining was the end. However, now, managers are looking to the analytical result as the beginning. Business users use the analytics to answer two critical questions. They want to know how they turn the information into action and what effect each action will have. It is not enough to simply know the results of the analytical process. Data mining used to discover an interesting group of customers. Now we need to move to action and measurement by saying, “Here are a group of people that should be presented with product X when they call inquiring about product Y. By following this course of action we project a 20% lift in call center revenues over the next 3 months.”

Our industry is excited about business analytics (i.e., the gathering and analyzing of business data in order to make better informed business decisions). However, in most companies there is a division of labor between the analytics (business) user and the analyst. Although the business user is an expert in his area it is most unlikely that the he is an expert in data analysis and statistics. The needs of both players need to be aligned. This is not an easy job.

For example, Kohavi, Rothleder, and Simoudis (2002) highlight the following challenges most organizations face as they try to create relevant, accurate, and timely analytics.

1. The time to crunch the numbers and analyze the data is never fast enough. But then, will it ever be? Will we just keep demanding faster and faster answers? When should we redefine “real-time” to “right time?”

2. Business users want to be a little more self sufficient. They want user-friendly interfaces that will allow them to rely less on other people to get the answers when they want them.

3. Data collection and analysis isn’t targeted. We want it all whenever we want it. We don’t take the time to define clear business goals and metrics. In the past, unrealistic expectations about data mining “magic” led to misguided efforts without clear goals and metrics.

4. We want analyze data that must be integrated from multiple sources. Most of the time we don’t have an efficient and cost effective manner to do this. The extract-transform-load (ETL) process is usually complex and when it is considered, the cost and difficulty are usually underestimated.

Most every company I work with is creating on some type of dashboard and/or scorecard. Both of these are delivered through some type of business analytics. It is a grueling process and I run up against each of the challenges mentioned above. I find that the biggest hurdle is to get the business user to define the business goals and metrics. When most people go down this path they believe, “If I can measure it, it must be important.” Many people become fascinated with “quantity” rather than quality. The “fun” side of seeing all these correlations and neat analysis paralyzes them. Rather than thinking of what action they will take and what effect that action will have, they enter into the common analytic stupor of “paralysis by analysis.”

Kohavi, R., Rothleder, N.J., Simoudis, E. (2002). Emerging trends in business analytics. Communications of the ACM, 45(8), 45-48. Retrieved June 22, 2008, from http://ai.stanford.edu/~ronnyk/cacmEmergingTrendsInBI.pdf