Those of you in charge of managing the multichannel marketing efforts of your companies need to keep “frozen” snapshots of your databases for future analysis. In other words, you should regularly create a complete copy of your database to save for future needs. Data storage is increasingly inexpensive, so when in doubt, make a copy of the data.
This includes keeping a table of exactly who was selected for specific promotions. And while your source key history is an integral part of your planning and a wonderful summary of the results of your marketing campaigns, there are times when you will wish you had more detail than the combined results of a segment of customers or prospects.
Here are some situations where having snapshots of your database and individual level marketing spending are very helpful:
* Lifetime value (LTV) calculations. These are essentially about matching downstream revenue to downstream marketing spending. Having customer-level data about exactly which promotions the customer has been selected to receive joined to a transaction table makes LTV calculations very easy for all kinds of customers, not just new ones. The promotion data should include customer ID, source key, and match code for all customers promoted.
* Reactivation. Most companies are working very hard to reactive old customers. Reactivated customers may have a very different “new” LTV based on their prior behavior, original source, the reactivation offer, or the offers they have responded to in the past. Having a promotion table tied to transactions makes selection and detailed comparison possible, and that knowledge will show the way to increased profits.
* Testing of selects. Circulation planning in the multichannel environment is, and should be, a constant process of refinement and testing. It is possible–indeed, likely–that at some point there will be a test that indicates a new variable that makes a marked difference between response and nonresponse. When you use that variable, you are changing course, and your previous source key results might not be a valid reference for planning current and future promotions. If you have copies of the database from previous periods, you can relatively easily resegment the historic database copies based on new variables, join them to the promotion table of who was selected for marketing, and regain the ability to use past promotions as reference for planning new ones. It can be a great way to confirm your new findings too.
* Regression analysis. A client with 100 stores and an eight-figure catalog and Internet enterprise in a highly seasonal business wanted to be more proactive in its marketing efforts. It was time to more accurately tie marketing expenditures to their revenue effect because there was investment available to grow the business, but how should they spend? The company needed to dig deep in customer behavior to find the combinations of variables that were most predictive of buying behavior.
The client elected to conduct CHAID analysis on its data. Source key descriptions indicated the selection criteria for different source keys, but lacking the frozen copies of the database at different times meant we couldn’t re-create who had actually received past promotions. This limited the analysis to the data from the current season, one that the company referred to as a shoulder season marketing campaign. It was extremely disappointing to find that the scoring method was not predictive in prime season. Being able to re-create who had received previous prime season promotions would have meant a one-time setup of datasets with more than one season’s promotions analyzed and a jump of one year in really applying the learning.
Remember that when you are building response models, having the ability to research, apply, and validate the model across multiple promotions helps the modeler more thoroughly understand the dynamics of variables and tune the model.
Duff Stokes is president of Rowayton, CT-based Contact Strategy, a direct marketing consultancy.