One of the best ways for database marketers to maximize their budgets and increase ROI is to find the appropriate mix of contact methods for each customer segment. While there’s no tried-and-true, one-size-fits-all strategy, testing can lead you to your optimal mix strategy.
When testing, there are many factors to consider.
- Control for seasonality. Test outcomes will likely vary by time of year, so it’s important to compare test groups within the same time period. The holiday season will behave differently than other times of the year. Customers have a reason to shop, so your base will be diluted with gift-givers who may not have long-term customer value.
- Compare similar groups. When comparing test cells, factor out elements that could cause a difference in response that is not associated with your test variables. If you’re testing the lift of email over direct mail, the direct mail-only group should also consist of contacts with an email address. Otherwise the direct mail and email group may have a different profile and propensity versus a contact group with land addresses only.
- Create proper sample sizes. Epsilon Targeting recommends aiming for 250 responses per test group to achieve 95% confidence in your results. However, you may need to decrease test group size due to cost or the available test universe.
For example, if you want to test the incremental impact of email and display ads on ROI versus using direct mail alone, ensure that all groups have an equivalent profile. In this case, groups should have a presence of cookies and email addresses, even if their treatment is to receive direct mail only. Because of the availability of contact information, there will naturally be more people who have only one component (such as a land address), compared to those who have all three (land address, email and cookie).
- Choose the right test method for your needs. Time-based testing measures the lift of one method versus another by measuring group performance over a period of time. Some tests may be better served by varying contact method by geographic similarity. “Paired market testing,” where two or more like geographic areas are identified and different strategies are executed to each simultaneously, can be helpful to reduce the length of a test or when randomization has proven to be difficult.
Karren Madson is the director of product management at Epsilon Targeting.
