Performance Max campaigns — what actually happened when we moved budget to them

Sometime in early 2023, the pressure to move clients to Performance Max became hard to ignore. Google was pushing it in every rep conversation. The platform was actively sunsetting Smart Shopping in ways that made PMax the path of least resistance. We were skeptical, but we ran the experiment properly. Here is what we found.

We moved three clients to PMax over a six-month window. One was an e-commerce brand selling kitchenware online. One was a service business in Pune running lead generation. One was a D2C skincare brand with a healthy catalogue. All three had stable campaign histories we could benchmark against. We tracked cost per conversion, conversion rate, impression share, and blended ROAS over the test period.

What PMax did well

For the kitchenware brand, PMax found converting audiences we had not tested. Performance stabilised after about four weeks of learning and ROAS held roughly steady with lower management overhead. The asset group structure forced us to write more creative variations than we had been running in standard Shopping, which turned out to be genuinely useful.

The skincare brand also responded well in the first eight weeks. We saw new customer acquisition picking up and the attribution looked reasonable given their purchase cycle.

Where it struggled

The lead gen client in Pune was a different story. PMax has effectively no search term transparency. For a service business where intent signals matter, that opacity was a real problem. We were seeing conversions but we could not tell which search queries were driving them, which meant we could not use that information to improve the account elsewhere. The CPA was acceptable but we had less confidence in the quality of leads we were sending to the client.

The other issue across all three accounts was brand cannibalization. PMax will serve against your own brand terms unless you explicitly exclude them as negative keywords, which requires a workaround because PMax does not natively support negatives the same way standard campaigns do. Once we sorted that, results improved, but it should not be a workaround. It should be a default setting.

Our current position is that PMax is a reasonable choice for e-commerce clients with good product feeds and catalogue depth. For service businesses dependent on search intent, we still prefer standard Search campaigns where we can see and control what is happening. The transparency trade-off is not worth it when a bad lead costs the client three hours of a salesperson's time.