We paused a PMax campaign running on zero setup, built the infrastructure, then rewrote the site from scratch

A mid sized car rental operator based in Baku, running a fleet of 50 vehicles. The target audience was tourists in Azerbaijan: visitors who had just arrived and were looking for a car, plus travellers whose need for a rental came up mid trip.

The difficulty is that this audience sits geographically on top of local users. Location targeting alone does not separate a tourist from a resident: both search in the same city at the same moment. The split can only be built on language, search behaviour and intent signals.

There was no measurement setup on the site at all. No GTM, no GA4, no conversion tag, no dataLayer. The campaign had been created, but targeting, signal and the conversion definition were left empty.

For Performance Max specifically this is the most expensive kind of mistake. PMax is not a targeting tool, it is an optimisation engine: it decides what to do from the conversion signal it receives, not from what you intended. If you do not tell the campaign which action is valuable, the algorithm defines “value” itself and in practice buys the cheapest impressions. The result: budget flows, clicks arrive, rentals do not.

In a business targeting tourists this compounds. Because the audience sits geographically on top of local users, the targeting filter has to be the conversion signal, not the location. With no signal, PMax picks the cheapest inventory, which in practice means broad, low intent traffic inside the country. The campaign was optimising for the exact opposite of the audience it needed to isolate.

In short: the only thing separating a tourist from a local user was the data itself, and that data was never being collected.

Three weeks of spend also left no usable data behind. Because conversions were not measured, that period cannot be analysed after the fact.

Once the right conversion signal started flowing, the campaign reached the target audience a little more accurately each week. Today the client can reach the tourist segment they aim for at a cost below their previous spend level.

The data from the first three weeks cannot be recovered. Because conversions were not measured, that period is not analysable after the fact, so the comparison runs only on the period after the new setup. This is the hidden cost of a campaign without a setup: the money goes, and no learning stays behind.

Showing improvement on an account with zero setup proves the effect of a correct setup. But we do not promise a jump of the same size on an account that is already set up correctly. For an account with a healthy setup and a problem elsewhere, the expectation should be different.