A Performance Max playbook that respects your margins

Performance Max is powerful and opaque in equal measure. Hand it a goal and a budget and it will optimize hard — but optimize toward what you told it, which is often not the same as profit. The brands that win with PMax are the ones that shape its inputs carefully instead of treating it as a magic box.

Optimize to value, not conversions

If you optimize for raw conversions, PMax will happily buy you the cheapest conversions it can find — frequently low-value ones. The fix is to feed it real value signals:

• Send conversion values, not just conversion events — ideally actual revenue or margin.
• Where possible, pass offline conversions back so the model learns which leads became customers.
• Use target ROAS once you have enough value data for it to be stable.

The model is only as good as the objective you give it. Define success as profit and PMax chases profit.

Control the inputs you still can

PMax removes a lot of manual levers, but not all of them. We use the ones that remain deliberately:

• Asset groups structured by theme or product line, each with strong, distinct creative.
• Audience signals built from your best first-party data to give the model a warm start.
• Account-level exclusions for brand terms and irrelevant placements so you're not paying to recapture existing demand.

Feed it, then get out of the way — mostly

Once the objective, signals and creative are right, PMax needs room to learn. Resist the urge to make daily changes that reset the learning phase. Instead, review on a steady cadence, refresh creative before fatigue sets in, and adjust targets as your data matures.

Run this way, Performance Max stops being a black box you fear and becomes a reliable engine for profitable demand capture. That balance — automation guided by human strategy — is exactly how we run search for our clients.