Can a new product split logic improve PMAX campaign performance?
The Problem
The products in the account were split into two PMAX campaigns: Heroes and Others.
However, the split didn’t take into account what matters most: how well each product actually converts. Only products with a high number of conversions made it into Heroes. Products with a good conversion rate but fewer conversions ended up in Others, even though their performance deserved more room.
We wanted to test whether changing the split logic to one based on conversion rate would help the campaigns perform more efficiently. We ran the test in market A, with market B serving as our control group.
The Solution
New product split logic
We kept the campaign structure but changed the segmentation rules.
The deciding factor is now the conversion rate over the last 120 days:
Heroes – products with a conversion rate of 1.5% or higher
Others – products with a conversion rate below 1.5%
Regular refresh
The split isn’t static. Labels are updated every 7 days, so a product that improves automatically moves into Heroes, and a product that loses performance moves to Others. This way, the campaigns always work with up-to-date data.
Results
After changing the split logic, we saw the following changes:
- Cost: +25 %
- Clicks: +29 %
- Revenue: +40 %
- ROAS: +12 %
Heroes vs. Others
ROAS improved similarly in both campaigns, but each got there in a different way:
Heroes – neither cost nor clicks grew significantly; the growth came mainly from revenue. Products with the best conversion rate managed to get more out of a similar budget.
Others – everything grew significantly here: cost, clicks, and revenue. The more precise split gave weaker products room to perform, which translated into revenue.
Control market
We made no changes to Market B, which served as a control group—yet results improved slightly. Revenue increased by 4.6% compared to the reference period.
This means that external factors, such as seasonality, market demand, or changes in Google’s algorithm, likely contributed to the growth. By comparing the results against the control market, we can conclude that the experiment had a positive impact.
Conclusion
Splitting products in PMAX campaigns by conversion rate brought performance growth and an interesting look at how strong and weak products behave when they’re split on different types of data.
