How modeling ad targeting after just your most qualified contacts impacts qualified lead generation results
NextAfter
Experiment Summary
Timeframe: 01/01/2023 - 01/31/2023
As a part of our recent experimentation with our own acquisition efforts, we decided to try specifically modeling a lookalike audience in Facebook (Meta) Advertising based upon our most engaged/best qualified contacts in the database, rather than our control approach (which is a mixture of interest-based targeting, lookalike models from our overall email database, and directly uploaded prospecting lists).
Would this improve the overall volume of the new qualified leads (and decrease the cost per qualified lead)?
Research Question
We believe that modeling after our most highly qualified contacts for our target audience will achieve an increase in qualified leads generated (and/or decrease our cost per new qualified lead generated) because it will be based upon a better source list.
Design
Results
| Treatment Name | Conv. Rate | Relative Difference | Confidence | |
|---|---|---|---|---|
| C: | Control | 2.0% | ||
| T1: | Treatment #1 | 1.1% | -44.1% | 76.1% |
This experiment has a required sample size of 1,497 in order to be valid. Unfortunately, the required sample size was not met and a level of confidence above 95% was not met so the experiment results are not valid.
Key Learnings
After a month of experimenting with this hypothesis, we were unable to fully validate this concept, although we have some directional data that suggests that it is not necessary to model your prospecting audience within Facebook (Meta) Advertising solely based upon your most qualified contacts.
It’s worth documenting some other movement on KPI’s as a part of this study, although it did not validate.
Those are:
- The control list had a 41% lower cost per qualified lead generated when compared to the treatment list.
- The control list generated 8% less “generic email domain” leads when compared to the treatment list.
- The control list did have a slightly higher cost per lead generated (+5.3%) when compared to the treatment list.
Conclusion:
Our hypothesis as to “why” this is that it could simply be due to pure audience size — whereas the “control” list was 41.9% larger than that of the “Qualified Contact Lookalike” list used within the control. This manifests itself in the CPM (or “Cost per Thousand Impressions Served”) metric — which was 11.9% less expensive for the larger control list when compared to the treatment.
With all of that said, this evidence suggests that it’s not worth the effort to model just off of your most qualified contacts within Facebook (Meta) Advertising, at least as it pertains to our own B2B marketing campaigns. It may differ within a donor program, and is worth further experimentation in that regard.