1) Design your experiment
Define your objective
Clearly state what you want to measure and why.
Define channels to test
Define which channels you want to test, for example Meta prospeting or Meta retargeting. The size of the channels affect statistical significance of the experiemtn - ensure you choose channel with sufficient spend.
Select participating geographic regions
Choose geographical regions that will participate in the test. Some of the geographies will be used as control, and some will receive a treatment. Choose geographic areas that are large enough to provide statistically significant data. They should have Marketing Mix Modeling implemented, so that you can compare your results to MMM.
Design marketing intervention / treatment
Different intervention options include for example hold-out tests or a change to the media spend, for example -50% or +100%. Testing different spend levels might reveal information about marginal ROI, but are more challenging to execute and require larger amount of participating regions, than a simpler hold-out test.
Define control markets and test markets
Assing, which pariticipating markets will acts as control markets, and which will receive the treatment.
Define test period & length
Choose a study period long enough to capture the campaign’s impact, accounting for any lag in consumer response
2) Implement the experiment
Conduct the experiment in accordance with the experiment design.
3) Analyse
Use a rigorous analysis methodology to analyse the test results. For Sellforte customers, this is included in the service.
4) Take action
Based on test results, you can
Validate your Marketing Mix Model: Are the experiment results in the range of what MMM is telling?
Calibrate your Marketing Mix Model: If your test results deviate from MMM results, start using this information in MMM calibration. Improved calibration for one channel improves the quality of the whole model