Getting More Bang for Your Marketing Buck

June 29, 2017   |   Vol. 16   |   Issue 6
By Elisa Joseph Anders, MGI Account Director

Getting More Bang for Your Marketing Buck

You’ve picked the low-hanging fruit in your housefile (your organization’s files of active and former members, buyers, and non-member non-buyers). You have repeatedly marketed to your strongest lists and gotten good results.

But you wonder about the house lists to which you don’t usually market. Could you get more mileage out of your housefile by mailing more of the prospects in it?

Yes. Through predictive modeling, you can identify prospects who are most likely to join or buy – across all your lists, not just your top-performing lists. A scored prospect model ranks all your housefile prospects, by decile, by their likelihood to join or buy. The result: more members and buyers, higher sales, and better ROI from your marketing campaigns.

We did this for one of our clients recently and achieved phenomenal results in our market test: the top three scored prospect deciles performed 94% better than our “house list control,”a random sample of records taken across all the lists in our client’s housefile. Our next steps: roll out to these top three deciles and test deeper into the deciles.

This approach is a powerful solution for our client, and it can be for you too. Predictive modeling enables you to market more frequently and productively to prospects most likely to join or buy, while eliminating wasteful resources spent on those least likely to join or buy.

So how does this work? Say you’re building a membership scored prospect model.

  1. Assemble all your house prospect lists. Don’t just include the obvious – e.g., lapsed members and book buyers – but be really complete. Find any lists of non-members who have bought from or engaged with your organization and whose records your organization has in its AMS or in files outside its AMS. Include paid and free journal and enewsletter subscribers, shopping cart abandons, white paper downloads, career center visitors, conference and webinar registrants, professional development and certification program buyers, chapter event attendees, and so on.
  2. Create a masterfile of all the prospects across these lists. This is a many-to-one mapping. For example, say Jane Prospect abandoned her online shopping cart while making a purchase last year, subscribes to your enewsletter, and attended a webinar last month. She would appear only once in the masterfile. Her record would be tagged with her three initial list sources.
  3. Build your scoring algorithm. Look at the prospects who became members during your previous campaigns to see what they have in common, and the prospects who didn’t become members to see what they have in common. By comparing campaign performance history, you can decide what to use as scoring elements in your model, and how heavily to weigh each of these elements. Some examples to consider: list response rate to the past year of completed direct marketing campaigns, frequency of appearance on multiple lists, SCF or zip code, dollar value of purchases, recency of purchase or engagement, demographic and firmographic data (e.g., title, years of experience, type of organization), presence of mail and email addresses, and so on. The combinations of the elements from which you can pull – and the weighting you give to each element – form your scoring algorithm.
  4. Build and run your model, applying the algorithm and assigning a score to each prospect in your masterfile. Create ten deciles – the top 10% of scored prospects in decile 1, the next 10% in decile 2, and so on.
  5. Validate your model in a pre-market test. Test the validity of your model against your prospect and new member files from historical campaigns. To do so, run these files through your model and assess the response of each decile. The model is validated if the top deciles perform better than lower deciles.
  6. Test your deciles in your next campaign, starting with the top three, against a randomized house list control group. The model is validated if the top deciles perform better than the control group.
  7. Continue to test deeper into your deciles based on prior decile testing results.
  8. Periodically update your lists and re-run your model to keep your decile lists fresh.

With a 94% lift in response from only the top three deciles, the impact on our client’s program is huge. Over the coming year, this should generate thousands of additional members and hundreds of thousands of dollars in additional dues and non-dues revenue. And it will save precious marketing dollars that can be redirected to other parts of the marketing program or to other needs in the organization.

Want to learn more about how scored prospect modeling works or how it can work for you? Please contact Elisa Joseph Anders at 703.706.0339 or at

Posted in: