Performance Max examples: what good actually looks like.
Abstract advice only goes so far. Three real setup patterns from accounts we run, and the reasoning behind every choice.
Most Performance Max advice is abstract — structure well, exclude brand, feed good signals. True, and hard to act on without seeing shapes. So here are three setup patterns from accounts we operate, anonymized but real, with the reasoning behind each choice made explicit. None of them is exotic. That is rather the point: good PMax is a set of deliberate, boring decisions compounding.
Example 1: high-SKU ecommerce retailer
A retailer with tens of thousands of SKUs across wide margin variance. The defining decision: campaigns split by margin tier, not by category. Three PMax campaigns — premium (high margin, target ROAS ~2.5x), core (mid margin, ~4x), and clearance (thin margin, ~6x) — because a blended target across mixed margins quietly starves the products that make the money. Within each campaign, asset groups follow the major categories with listing groups scoped to match. Brand is excluded everywhere and lives in its own Search campaign.
The other half of the work never touches the campaign UI: the feed. Titles rewritten to lead with what customers search (attribute-rich, brand-last for unbranded categories), GTINs complete, availability synced multiple times daily so the campaign never spends against out-of-stock items. With this many SKUs, the feed is the campaign. Channel-level reporting — now that Google exposes it — confirmed what the structure assumed: the overwhelming majority of converting spend runs through Shopping inventory, so feed work compounds hardest.
Example 2: considered-purchase brand
A brand selling a four-figure product with a research cycle measured in weeks. The temptation with long cycles is to judge PMax on last-click within the window, which undercounts it badly. The structure here: one PMax campaign, three asset groups split by entry point into the research journey — problem-aware ("is X worth it"), comparison ("X vs Y"), and ready-to-buy — each with genuinely different creative and its own landing environment. Search themes are used deliberately per group, and the recently expanded theme limits get used in full rather than left half-empty.
Measurement is where this account earns its result: enhanced conversions plus cart data, a micro-conversion (configured quote) valued at a fraction of a sale to give the algorithm intermediate signal across the long cycle, and a monthly geo-split sanity check on incrementality. The campaign target is set against the micro-to-sale ratio measured quarterly. Without that scaffolding, the same structure would drift toward whatever converts within seven days — which for this brand is the wrong customer.
Example 3: lead gen with offline import
A services business where a form fill is worth anywhere from nothing to five figures. The non-negotiable foundation: offline conversion import from the CRM, pushing qualified-lead and closed-won events back into Google with values attached. PMax bids toward qualified pipeline, not raw form fills. Campaign-level negative keywords do real work here — job-seeker terms, free-intent terms, geographies the business cannot serve — and the search terms reporting now available for PMax gets reviewed on a fixed cadence, feeding the negative list the way a Search account would.
The pattern across all three: the campaign structure is the smallest part. Feed quality, conversion signal quality, and honest measurement decide whether PMax compounds or corrodes. Structure just decides how legible the results are.
What all three have in common
- Brand excluded, always, with a dedicated brand Search campaign catching that traffic at honest prices.
- One change per learning cycle, annotated, with budget moves capped around 20% per step.
- Judged on account-level incrementality — total conversions and blended return — not campaign-reported ROAS.
- Creative refreshed on a cadence tied to fatigue signals rather than the calendar.
- A conversion signal the team actually trusts, because every downstream decision inherits its quality.
If your setup looks nothing like these, the gap is usually not talent — it is that nobody wrote down the reasoning. Our PMax control guide covers the full lever-by-lever playbook, and if you want a second pair of eyes on your own structure, a free audit is the fastest way to get one.
Should Performance Max campaigns be split by profit margin?
For retailers with wide margin variance, yes — it is often the single most consequential structural decision. A blended target across mixed margins quietly starves the products that actually make money, because the system optimizes toward the average. Splitting into premium, core and clearance campaigns, each with a target ROAS matched to its margin band, lets every tier be bid at a rate its economics can support.
How do you run Performance Max for a long consideration cycle?
Give the algorithm intermediate signal and stop judging it inside a seven-day window. In practice that means asset groups split by entry point into the research journey — problem-aware, comparison, ready-to-buy — each with genuinely different creative and landing environment, plus a valued micro-conversion such as a configured quote priced at a fraction of a sale. Set the campaign target against the micro-to-sale ratio and re-measure it quarterly.
Can Performance Max work for lead generation?
Yes, but only with offline conversion import as the foundation. Push qualified-lead and closed-won events back into Google with values attached so the campaign bids toward pipeline rather than raw form fills, which are worth anywhere from nothing to five figures. Campaign-level negative keywords do real work here too — job-seeker terms, free-intent terms, and geographies you cannot serve — fed by reviewing the PMax search terms report on a fixed cadence.
What do well-run Performance Max accounts have in common?
Five habits, none of them exotic: brand excluded with a dedicated brand Search campaign catching that traffic at honest prices; one change per learning cycle, annotated, with budget moves capped around 20 percent per step; judgment on account-level incrementality rather than campaign-reported ROAS; creative refreshed against fatigue signals rather than the calendar; and a conversion signal the team genuinely trusts, since every downstream decision inherits its quality.
Written by The ADSRUNNER team. If this resonated and you want to apply it to your own account, you can book a strategy call or run a free audit.
How we research, source figures, and handle corrections: editorial policy.