Performance Max took bids, keywords and placements away from marketers and gave them a single target slider and reports that show less than we would like. The campaign can still be steered; the levers are simply elsewhere: in the feed, in margin and stock data, and in what you feed the algorithm. This guide shows what you can see in PMax, what you cannot, and how to keep your hands on the wheel regardless.
Mechanics
How Performance Max differs from standard campaigns
Definition
Performance Max is a Google Ads campaign type in which one set of assets (feed products, text, images, video) serves across all Google channels at once: Search, Shopping, Display, YouTube, Gmail and Maps. You set the target and budget; the algorithm decides on bids, placements and product selection. Control shifts from campaign settings to the quality of the input data.
The easiest way to understand PMax is as a machine with three inputs and three outputs:
What you provide
Performance Max
The algorithm chooses bids, channels, queries and products itself to meet the set target.
What you get
The takeaway from this diagram is practical: since input number one is the feed, the quality of titles, prices, availability and labels determines what the algorithm can learn at all. The basics are covered in the guides to the product feed and Google feed fields.
Visibility
What you can see in Performance Max and what stays hidden
Frustration with PMax comes from the gap between expectations and reports. An honest list looks like this:
What you will see
- Campaign results: cost, conversions and value calculated according to Google’s rules
- Per-product view: impressions, clicks, cost and conversions for feed products
- Listing groups: results of subdivisions by labels, brand, type
- Search categories: a general view of search themes
- Asset ratings: which text and images the algorithm considers effective
What you will not see
- Full search terms with cost per term, as in Search campaigns
- Cost split by channel (how much went to YouTube, how much to Shopping)
- Bids and auctions: why the algorithm bought this click and not another
- Margin and stock: the campaign knows neither your cost of goods nor your stock levels
- Real orders: conversions are figures based on Google’s windows and modelling
PMax hides the mechanics but leaves the results in plain sight: orders, margin and stock live in your shop. That is where control returns.
Assessment
How to assess PMax on orders and margin
Since the campaign report tells half the truth, assessment moves to shop data. Three control questions, built on whole-shop paid orders and margin set side by side with campaign data:
First: does the shop earn more since PMax started spending more? MER (revenue divided by total spend) and profit after ad costs in weekly windows answer this more honestly than campaign ROAS, which depends on windows and modelling. Second: what does PMax actually sell? The per-product view set against margin shows whether the budget goes to profit leaders or to high-volume products with no margin. Third: is the growth just your own brand sales relabelled? A comparison with brand traffic and whole-shop sales shows how much the campaign adds and how much it merely puts its name on.
A note on methodology: PMax conversions follow the same rules as the rest of the ad dashboards (attribution windows, modelling, no order statuses), and product-level cost refers to the product that was clicked, not the contents of the basket. How to read such figures is explained in the guide to conversions and attribution.
Control
Control without bids: feed, labels and exclusions
PMax offers four control levers, and all of them sit outside bid settings:
| Lever | How it works in practice |
|---|---|
| Campaign structure | separate campaigns for segments with different goals (margin leaders, long tail, seasonal), split by feed labels |
| Product groups and exclusions | only the right groups stay in the campaign; low margin, low stock and end-of-line items are removed deliberately |
| Feed quality | titles using customers’ own words, correct prices and availability: this is the only language in which you tell the algorithm about your products |
| Target matched to the group’s margin | a group with a 45% margin can tolerate a lower target ROAS than a group with an 18% margin; one shared target averages them out and harms both |
The stock and margin labels that make this possible are built from shop data and sent back to Merchant Center through a supplemental feed; the mechanics are described in the guides on margin and stock levels.
Real risk
PMax with a single target across the whole catalogue optimises an average that does not exist. The algorithm will hit the target by mixing products with 45% and 15% margins; the campaign report will look green, while the shop’s profit is a matter of chance. Without a split by margin nobody will notice, because the campaign does not know the cost of goods.
Vigilance
Warning signs in a PMax campaign
A few patterns recur often enough to be worth checking every week:
Rising cost with flat shop sales: the campaign looks great, MER falls; a classic sign of averaging or of double-counting branded sales. Budget drifting to the long tail: in the per-product view, the share of products outside the leaders grows; check whether groups and labels still match the data. Products disappearing from serving: a sudden drop in impressions for specific items is usually a Merchant Center diagnostics issue or zero stock, not an algorithm decision. Results jumping after every change: adjusting the target and structure too often keeps the campaign in a permanent learning phase.
Key takeaway
The rhythm of working with PMax: infrequent changes, frequent observation. Make target and structure changes one at a time and in larger steps, evaluate them in weekly windows based on orders, and leave day-to-day vigilance to signals: MER, budget share per group and product serving.
DataOrganizer
How DataOrganizer helps control PMax
All the figures PMax will not show you are in the connected data model. Every day, DataOrganizer puts campaign spend and results (including the per-product view from product campaigns) side by side with the shop’s paid orders, margin per SKU and stock levels. On top of that come margin and stock labels, which return to Merchant Center via a spreadsheet and shape the campaign structure.
PMax spend structure against margin (furniture shop, illustrative data):
| Product group | Share of spend | Share of conversion value (Google Ads) | Avg. margin |
|---|---|---|---|
| Leaders (high-margin) | 31% | 52% | 44% |
| Fast movers (low-margin) | 47% | 36% | 17% |
| Long tail | 22% | 12% | 31% |
Almost half of the budget is working on a group with a 17% margin. Recommendation: move it into a separate campaign with a conservative target and shift the freed-up budget to the leaders. The feed labels are up to date, so the change comes down to campaign structure.
FAQ
Frequently asked questions about Performance Max
Does Performance Max show search terms?
To a limited extent: you get insight into search terms or their categories, but without the detail known from Search campaigns. This makes account-level exclusions and work on feed titles all the more important, as they govern how products are matched to queries; more in the guide to search terms.
How do you exclude products from a Performance Max campaign?
Through listing groups: only selected groups stay in the campaign (ideally based on custom labels from the feed), and the rest are left out. This keeps low-margin, low-stock and end-of-line products out of the campaign without manually excluding individual SKUs.
Does Performance Max claim credit for brand sales?
It can: PMax covers Search, so some conversions come from queries containing your shop’s name that would have ended in a purchase anyway. You can gauge the scale by excluding branded terms or by comparing campaign results with total shop sales and branded traffic, rather than with the campaign report alone.
How many Performance Max campaigns should a shop run?
As many as you have genuinely different goals: usually two or three (margin leaders, a conservative long-tail campaign, a seasonal one for the peak). A data-driven split works better than an intuitive one, and splitting too finely limits conversions per campaign and lengthens learning.
How long does Performance Max take to learn after changes?
It depends on conversion volume: from a few days to a few weeks. The practical rule: change less often and in larger steps, one change at a time, and evaluate in weekly windows based on shop orders. Several changes at once make it impossible to separate their effects.
Cheat sheet for the whole guide
Three principles for controlling Performance Max
1. Control lives in the inputs. The feed (titles, prices, availability), margin and stock labels, and a segment-based campaign structure replace bids and keywords. The quality of these inputs sets the ceiling for results.
2. Evaluate on orders, not on the campaign report. MER and profit after costs in weekly windows, the per-product view matched with margin, and a brand test reveal the truth the PMax report cannot access, because it knows neither your cost of goods nor your order statuses.
3. Infrequent changes, frequent observation. Single, larger adjustments and a weekly evaluation rhythm instead of tweaking sliders every day. DataOrganizer keeps all the necessary reports in one model: campaign spend, orders, margins, stock levels and labels ready for the spreadsheet.
DataOrganizer · PMax measured by profit
Connect Google Ads and your shop: campaign spend sits next to shop orders, margin and stock levels in one model, and labels for splitting campaigns go back to Merchant Center via a feed sheet. Demo account available instantly; you pay only for the sources you connect.