In a recent MDM survey, we asked distributors how they plan to grow revenue in 2015. Just behind product line/category expansion was growing revenue from existing customers, or increasing wallet share. There are many ways to model market and wallet share. So I thought it timely to share some research and programs a pioneer in wallet-share definition – IBM – pursued to build effective models around estimating information technology expenditures in their customers and prospects.
IBM threw a lot of analytics brainpower at this; you might pick up a few ideas from a review of what they did. The project started in the early 2000s when IBM’s Predictive Modeling Group evaluated ways to model opportunity – both at a customer account level as well as the larger market. Their research focused on two analytic models (Nearest Neighbor and quantile regression) that they found most successful. But it went deeper than that.
The company wanted to be able to focus sales resources on high-potential accounts and redeploy sales effort currently in “harvest” accounts – those with little upside potential – to those with a bigger opportunity. To do that, they had to find a way to calculate account potential. Share-of-wallet could also be used as a performance metric.
One of the key takeaways is that they landed on a hybrid analytics model that combines top-down and bottom-up modeling approaches. The top-down model starts with an aggregate total for a product category and then segments it across markets and accounts, modeling it proportionally. This is what MDM Analytics does using Dun & Bradstreet employee counts at specific accounts to model industrial product demand by category.
The bottom-up approach builds a predictive model based on customer transaction data. One way is to calculate wallet information from customer data, then build a model to predict wallet at other customers and prospects. We have built some of these hybrid models with Real Results Marketing for mutual clients with some solid success.
After refining the methodology from about 2002-2007, IBM rolled it out globally in a market alignment program. At that point they went pretty quiet, but the little bit they shared indicated it had a revenue impact of more than 10 percent for those sales groups that put it to work. That’s significant. The fact they rolled it out globally is an indicator that the analytics were working – and I imagine quite well.
This was a fascinating bit of work that IBM shared publicly. I highly recommend reviewing the IBM documents. And we are always interested in discussing how to adapt market analytic models to get better market intelligence.
For a slide deck and paper that outlines the IBM methodology, their implementation and success, as well as details on regression models used, email Kristen Gawalis at email@example.com.