As I talk to industry colleagues, I frequently hear the lament that prospecting and the cooperative databases do not perform like they used to. The common response is to simply reduce prospecting until cost-of-customer-acquisition goals are realized. Unfortunately, this also results in fewer new customers being added to the buyer file and, eventually, significantly fewer sales. While recent blogs have discussed new customer acquisition tools such as NaviStone, I want to focus on a technique we are testing with clients to improve prospecting with cooperative databases.
In past blogs, I reviewed several approaches for improving cooperative database results. These included the following:
- Prior to modeling, provide the co-ops with your full, up-to-date buyer file so that they do not return your own buyers and can optimize their modeling;
- Order the best models from at least three co-ops; each of co-ops will provide incremental performing names allowing you to “cherry pick” the best names from each co-op; going deep with a single coop generally provides inferior response;
- Create micro-segmentation in the merge that identifies multi names between each of the co-ops and any outside lists that you rent/exchange; this provides visibility to what portion of a model is really driving the performance (perhaps only the multis) and what portion should be discarded;
- Finally, perform “look back” analysis on various geo- and demographic variables over multiple drops to identify predictive attributes that can be applied as a precondition for future models. For example, with one client, we analyzed all prospecting over a three-year period by zip code finding large numbers of catalogs that had been mailed into zip codes performing at 20% of the response rate of catalogs mailed into an equal number of the best zip codes; these poor performing zip codes were suppressed from future models.
In this blog, I add another technique to this list that you should test. We have always known that co-ops send you the same prospects over and over again. This is generally positive because we also know from tests that the names that are “fresh” do not perform as well as the previously delivered names. We wanted to explore if there are records that have been mailed repeatedly without response and should be placed in a “non-responders” file, suppressing them from some or all future mailings.
For a client in the home décor category that mails monthly, we analyzed all prospecting mailings for a 24-month period calculating response rates for those mailed one time, up to 22+ times. The chart below shows the sales per piece for each group.
As expected, those that were mailed once had a slightly lower response than those that were mailed up to seven times. Response rates held steady through the 13th mailing. However, those contacted 14, 15, and 16 times fell below an acceptable response level; those mailed 17+ times “fell off the cliff.” There were 65,256 individuals in the 14-16 category that had received 605,609 prospect pieces in 2017; there were 71,793 individuals in the 17+ category that had received 789,185 pieces.
We conjectured that these non-responsive individuals did not decorate with the type of merchandise offered by our client. Others may have completed their home decorating and at this point would require additional time to pass before they would be in the market again; this however would not account for the fact that the coops over the 21-month period considered them top prospects due to their purchase activity! Irrespective of reason, we developed a strategy to cut our losses!
We immediately developed a non-responder suppress file of both of these groups that is updated monthly and sent to the coops. However, the 14 to 16 contact group will not be suppressed from modeling twice a year; the 17+ group will not be suppressed once per year and is mailed what is considered the best prospecting catalog of the year. We are eager to see how both of these groups respond to the reduced contacts. If they do not respond as a group to these reduced mailings, they will be permanently suppressed.
The suppressed group represented 18% of all pieces mailed during the period. We are optimistic that by suppressing the group, we can move these mailings to more productive model segments there by increasing overall co-op response. As with all strategies, continued testing will be necessary to see if our initial conclusions prove successful over time.
If you feel that your marketing strategy is in need of a tune up, email me at email@example.com