What Performance Max is actually for in 2026 — and where a Search campaign still wins

Performance Max spent its first years being argued about for one reason: nobody could see inside it. You gave Google a budget, some assets and a goal, and got back a number. Whether that number came from Search, YouTube, Gmail or a Display placement was Google’s business, not yours.

That argument is now largely over, and a lot of the advice still circulating is arguing against a version of PMax that no longer exists.

What has actually changed

Channel-level reporting is live. With Google Ads API v23, released on 28 January 2026, PMax performance breaks out by Search, Search Partners, YouTube, Display, Discover, Gmail and Maps, where queries previously returned a single generic MIXED value (Google Ads Developer Blog).

Two limitations matter in practice: channel data only exists from 1 June 2025 onwards, so you cannot analyse anything older, and asset-group-level channel reporting is API-only — it does not appear in the Google Ads interface at all (Search Engine Land). If you are working from the UI alone, you are seeing a coarser picture than the data allows.

More controls have been announced. Google has said first-party audience exclusions, budget projections, expanded demographic reporting, and placement reports segmented by network are coming, without committing to dates (Google Business). Check which of these are actually live in your own account before planning around them — announcement and availability are not the same thing.

Channel steering is in alpha. Google is testing a “Channels” setting that lets advertisers make positive or negative adjustments per channel. It isn’t fixed budget allocation: a positive adjustment relaxes the CPA threshold for that channel, a negative one tightens it (Search Engine Land, August 2026). Worth knowing about; not worth building a strategy on until it leaves alpha.

What PMax is genuinely good at

Finding demand you didn’t know to target. This is the real product. PMax’s advantage is breadth — it surfaces audiences, placements and query patterns nobody would have built a campaign around. With a wide catalogue and incomplete knowledge of who buys what, that breadth is worth money.

Scaling e-commerce with a healthy feed. Shopping inventory plus PMax is the combination the campaign type was built for. A well-structured, well-attributed product feed is the single biggest input to whether it works.

Consolidating a fragmented account. Forty campaigns sharing twelve conversions a month means the algorithm is starving. Consolidation into fewer, better-fed campaigns usually beats granularity, and PMax is one legitimate way to consolidate.

Efficiency at volume. Given enough conversions, PMax’s bidding is hard to beat manually. The question is always whether you have enough.

Where another campaign type still wins

Brand terms. Keep these out of PMax. Brand traffic is cheap, converts at rates that flatter any blended average, and will make a PMax campaign look like it’s doing something it isn’t. Run brand in its own Search campaign, exclude it with campaign-level negatives, and judge everything else on what’s left. In my experience this is the single most common reason a PMax campaign appears to outperform.

Anything where the exact query matters. Regulated categories, or products where a near-miss query means an expensive worthless click. Search with tight match types and real negative lists does this better, because control is the point.

Accounts with thin conversion volume. PMax needs signal to learn. Below roughly 30 conversions a month — a working threshold from my own accounts, not an official Google figure — you have given up control in exchange for very little. Start with Search, build volume, then consider consolidating.

Sharply uneven margins across the catalogue. PMax optimises to the goal you give it. If a 4% margin product and a 40% margin product both count as “a conversion”, you get more of the wrong one. Either split by margin band into separate campaigns with separate targets, or pass true margin as your conversion value. The second is better and harder.

When you need to know why. Even with channel reporting, PMax tells you where far better than why. During a diagnostic period — a performance collapse, a new market, a repositioning — the controllability of Search and standard Shopping is worth more than PMax’s efficiency.

Four things to actually do

Exclude brand. Campaign-level negative keywords exist. Use them, then look at your numbers again.
Feed before assets. Time spent on titles, attributes, categorisation and image quality returns more than time spent on headlines. Consistently true, consistently skipped.
Pull the channel data via the API. The asset-group breakdown does not exist in the interface. If a meaningful share of spend sits in Display or Gmail with nothing behind it, you now have evidence rather than suspicion.
Give it a target it can’t game. Feed real profit as your conversion value if you possibly can. A tROAS target on revenue optimises for turnover, and turnover is not the business you’re in.

The honest summary

PMax is now a reasonable default for e-commerce accounts with a decent feed and real conversion volume, and it was not three years ago. That’s a genuine change and it’s fine to say so.

It is still the wrong tool for brand traffic, for thin accounts, for regulated categories, for catalogues with wildly uneven margins, and for any period where you need to understand a problem rather than optimise past it. The campaign type is not the strategy. Where your money makes profit is the strategy.

If PMax is eating your budget and you can’t tell whether it’s earning it, that’s a diagnosable question — usually in days rather than months.