August 20, 2009

OMS: Real-Time Scoring

In my OMS projects, I create three different families of scores:
  1. Probability of a customer purchasing in the next twelve months.
  2. Amount a purchaser will spend in the next twelve months.
  3. Factor analysis scores for channel x merchandise combinations.

The scores yield a series of segments --- as few as maybe 80, as many as several thousand.

Each segment has a predicted future value --- that future value fuels the simulation of five year sales and profitability of your online business.

This information could easily be incorporated into your Web Analytics platform, allowing you in real-time or near real-time to see how today's purchasers influence the future trajectory of your business.

For example, you query customers who purchased in the twelve months ending last night. Each customer is scored via the OMS algorithm, and placed in one of the segments. Then, the future value for the next twelve months is summed across segments. At this point, you know that your current twelve-month buyer file is going to generate, say, $33,000,000 in the next year.

Tomorrow you have an e-mail campaign that goes absolutely bonkers. Woo-hoo! Replicate the scoring process mentioned above, just shift the twelve-month window by a day, and re-score everybody. Sum future volume. Say the total is $34,000,000. Now you know that your e-mail campaign went bonkers today, but also added a million dollars of value to your future business.

That's something you'd want to know, wouldn't you? And knowing it in real-time or near real-time is even better.

That's the power of the OMS algorithm.


Finance folks have ways to forecast that a recession is coming. In recent weeks, the topic of the "yield curve" suggests that the...