What Performance Max controls and what it decides itself
Google's documentation describes Performance Max as goal-based: the advertiser sets conversion goals, a budget and a bid strategy, supplies assets and optionally a product feedA structured file of a store's products, prices and availability used to build shopping and catalogue ads., and the system decides which inventory to buy and which combination of assets to show. The list of things the advertiser does not control is long by design — placement by channel, bid per auction, which headline pairs with which image, and which search queries trigger product ads. The list of things the advertiser does control is what this guide is about.
| The advertiser decides | The system decides |
|---|---|
| Campaign goals and the conversion actions that count | Which channel each impression is bought on |
| Budget and bid strategy with optional targets | The bid in each auction |
| Asset groups: the text, images, videos and logos supplied | Which assets are combined and shown where |
| Listing groups: which products from the feed are included | Which product shows for which search |
| Audience signals and search themes (as hints) | Who actually sees the ads |
| Brand exclusions, account-level placement and content exclusions, final URL expansion settings | Landing page choice within the allowed URLs |
| Location, language and ad schedule | Device and time allocation within those bounds |
| New customer acquisition goal setting | How existing customers are treated within that setting |
That split means the usual levers of a Search account — adding keywords, raising a bid, pausing a placement — mostly do not exist here. Influence is exercised before launch (structure, assets, goals) and through the few documented controls after it. Advertisers who try to manage Performance Max daily tend to reset learning repeatedly; advertisers who structure it well and then change little usually get the better result. The author's Google Ads expertise page describes how this fits the wider account approach.
How many campaigns
Performance Max learns per campaign, and a target applies per campaign. The decision of how many to run is therefore a trade between control (separate targets, separate budgets) and data (enough conversions per campaign for the bid strategy to work with). The default should be one campaign per distinct economic goal, and no more.
| Reason to split | Split? | Why |
|---|---|---|
| Different margins needing different ROAS targets | Yes | One target cannot serve both without under- or over-bidding one set |
| Sales versus lead generation goals | Yes | Different conversion actions and values; different bidding |
| Feed segment with its own budget rule (clearance, launches, seasonal) | Yes | Budget protection and a target that fits the segment |
| Different countries or languages | Usually | Location and language are campaign settings; currencies and prices differ |
| Different product categories with similar margins and goals | No | Listing groups and asset groups handle it within one campaign |
| Wanting to "see" performance by category | No | Use listing groups and product-level reports instead of splitting |
| A segment with too few conversions to learn on its own | No | Keep it in the larger campaign; a starved campaign bids badly |
The rule of thumb that follows is: split by target, not by taxonomy. If the result of a split would be a campaign with only a handful of conversions a month, the split costs more in learning than it returns in control; merge it back and use listing groups and labels for visibility instead. The product feeds guide covers the custom labels on which margin and segment splits are usually built.
Asset groups and listing groups
An asset groupThe Performance Max unit holding the text, images, videos and signals for one theme, product set or audience. is a set of creative assets (headlines, long headlines, descriptions, images, logos, videos, a business name and a final URL) organised around a theme, together with optional audience signalsOptional audiences, such as customer lists, attached to a Performance Max asset group as a starting hint for the system rather than a targeting limit.. A campaign holds one or more asset groups, and Google assembles ads from the assets within a group. Group by message, not by product line for its own sake: an asset group for a category where the images, headlines and landing page belong together, a separate one where the message changes. Asset groups within a campaign share the campaign's budget and target; they are not a budget control.
Listing groups decide which products from the feed an asset group can advertise. They are built by subdividing on feed attributes — product type, brand, category, custom labels, item ID or condition — and excluding what should not serve. By default all products are included; the structure work is excluding and subdividing. A listing group split by custom label is how a margin-based campaign structure is implemented: the high-margin campaign includes the "high" label and excludes everything else, the low-margin campaign does the reverse, and a product is in exactly one place. Because products inside one campaign share its budget, the shopping feed title test is a design for trying new feed titles on a random half of a category while reading the untouched half for spend moving towards the retitled products.
Asset quality matters because Google rates it and because it is the creative. Supply the full range of image sizes, several videos (Google documents that it may generate a video from other assets when none is supplied, which is a reason to supply one), headlines that work in any combination, and a logo. The asset reporting rates each asset and shows which combinations serve most, and low-rated assets should be replaced rather than left in place. The asset group tutorial builds one group from a written brief to its first asset report, step by step.
Audience signals and search themes
Audience signals are the most misunderstood control in Performance Max. Google documents them as suggestions: audiences (customer lists, remarketing lists, custom segments, interests, demographics) that tell the campaign where converting users are likely to be found so that learning starts faster. The campaign is not restricted to them and is expected to serve beyond them. A signal built from the best customer list is a good start; a signal built from a broad interest adds little.
Search themes are the search-side equivalent: phrases describing what the business offers, added to an asset group, that help the campaign find relevant searches it might not infer from assets and the landing page. They are documented as supplementary to what the system learns from the URL and assets, and Google notes they are treated with similar priority to phrase and broad match keywords in Search campaigns when deciding between campaigns. Use them for products or services the landing page describes poorly, for new lines without history, and for terms that matter commercially; do not expect them to behave as keywords with match types.
Neither control is a targeting restriction, and neither is reported as a segment in the way a keyword or an audience in a Search campaign is. Their value is in the first weeks of a campaign or asset group; after that the campaign's own conversion data dominates.
Brand exclusions, placement exclusions, URL expansion
The documented hard controls are few, and each matters.
Brand exclusions. A brand list attached to the campaign prevents it from serving on searches and placements associated with those brands. The usual use is the advertiser's own brand, so that brand searches are left to a Search campaign where the terms can be managed and the spend measured honestly. Competitor brands can be excluded the same way where the advertiser does not want to pay for them.
Negative keywords. Account-level negative keywords apply to Performance Max, and campaign-level negative keywords for Performance Max have been rolling out; the current help page states what applies to a given account. The negative keywords guide covers how to build and review lists; for Performance Max the lists are the main query-level hygiene available.
Placement and content exclusions. Account-level placement exclusionLists of websites, apps, video channels or app categories where an advertiser's ads must not appear. lists and content suitability settings apply to Performance Max's Display and video inventory. Campaign-level placement exclusions in the Search-campaign sense are not available; the account-level list is the tool, and the placement report (below) feeds it.
Final URL expansion. By default the campaign may send clicks to any page on the site it judges more relevant than the final URL supplied, and may use page content to match searches. This can be switched off so that only the asset group's final URL is used, or kept on with URL exclusions for pages that should never be landing pages (careers, support, account pages, blog archives). For product campaigns, feed links are used for product ads regardless; expansion affects the non-product traffic.
Customer acquisition goal. The campaign can be set to bid higher for new customers, or to bid only for new customers, based on customer lists and detected purchasers. This is the only documented way to shape whether the campaign spends on existing customers, and it depends on a customer list being maintained.
Feed-only vs full asset groups
A Performance Max campaign linked to Merchant Center can run with asset groups that contain only the product feed and no text, image or video assets. In that configuration the campaign serves product ads across Shopping and the surfaces where product listings appear, and does not build the display, video and text formats that need assets. Advertisers who want behaviour close to a Standard Shopping campaign, with Performance Max's bidding and precedence, use this.
Full asset groups extend the campaign to every surface and format. The trade is reach for control: more inventory, more conversions at the margin, and more spend on placements that are harder to evaluate, with view-through and cross-channel AttributionAssigning conversion credit to touchpoints. entering the picture. Google does not document the share of spend that goes to each channel for a given configuration; the channel reporting described below is the only way to see it.
A practical sequence many accounts follow: start feed-only where tracking is new or the brand is cautious about display placements, add assets once product performance is stable and the exclusions are in place, and compare account-level results across the change rather than campaign-level ones.
Splitting one campaign into two by margin
An online furniture store runs a single Performance Max campaign across 1,800 products with a target ROAS of 400%. Product-level reports show that sofas (gross margin 45%) return a ROAS of 620%, while accessories (gross margin 22%) return 310% and absorb 38% of spend. The break-even ROAS for accessories is about 455%, so the accessories traffic is being bought at a loss while sofas are under-served. The team labels products by margin band in the feed, builds two campaigns with listing groups on that label, and sets targets of 320% for sofas and 480% for accessories. After eight weeks, combined spend is unchanged at £24,000 a month, sofa revenue rises from £71,000 to £89,000, accessories spend falls to 19% of the total, and blended profit after ad cost improves by roughly £6,500 a month. Conversions per campaign remain above 60 a month in both, enough for the bid strategies to keep learning.
Reading Performance Max reports
Performance Max reports less than a Search campaign, but more than it did at launch, and the reports it has are the only way to see what the system decided.
Insights. Search categories and search terms insights show the groups of queries that drove product and search traffic, with conversion data at the category level. This is where brand terms, irrelevant categories and new demand show up, and it is the input to brand exclusions and negatives.
Asset group and asset reports. Performance by asset group, and ratings and combinations per asset. Use these to replace weak assets and to see whether a theme deserves its own group.
Listing group and product reports. Performance by listing group and by product, including which products received no impressions. For a product campaign this is the primary report; it answers which part of the catalogue the campaign is actually working.
Channel performance. Reporting that breaks spend and results down by channel (Search, Shopping, Display, YouTube and others) has been rolling out; where it is available it answers the "where did the money go" question that feed-only versus full asset group decisions depend on. Check the current help page for what the account can see.
Placement report. The placements where ads showed, used to populate account-level exclusion lists.
Report on ROASRevenue generated per unit of ad spend. and CPACost of each conversion. at the campaign level against the margin-derived targets — the break-even calculator gives the Break-even ROASMinimum ROAS to avoid a loss. that the target must clear, and the ROAS calculator the actual figure — and resist reading asset group figures as if they were ad group figures, because the campaign allocates between them and a weak-looking asset group may be weak because the system has stopped using it. Once the structure is in place, the Shopping and Performance Max playbook runs it week to week, with owners for the feed, product decisions, asset refreshes and uplift tests.
Running alongside Search and Shopping
Google documents two precedence rules. Within an account, Performance Max takes priority over Standard Shopping for the same products: if both are eligible, the Performance Max campaign serves. Between Performance Max and Search, an exact-match keyword that is identical to the query takes priority; in other cases, including phrase and broad match and search themes, the campaign with the higher Ad Rank serves.
Structure follows from those rules. Keep a Search campaign with exact-match keywords for the queries that matter most, including brand, and exclude the brand from Performance Max; the Search campaign keeps those queries and reports them properly. Where Standard Shopping is kept for specific products, exclude those products from the Performance Max listing groups explicitly, because otherwise Performance Max will take them whatever the Standard Shopping bids say. Where both run on overlapping products deliberately, accept that Standard Shopping will serve mainly what Performance Max does not want. The Performance Max launch checklist asks for that overlap to be recorded, alongside the goals, feed, asset group and brand checks, before a campaign is enabled.
Measure overlap at the account level. Campaign-level conversions move between campaigns when Performance Max launches, and a Search campaign that looks worse may simply have handed queries to Performance Max. The honest test is total conversions and total cost before and after, with the Smart Bidding guide covering how targets behave during that change. The hub page sets out the Standard Shopping versus Performance Max decision that this section assumes has been made. Where a before-and-after comparison is too noisy to trust, the Performance Max uplift test runs the launch in matched regions and reads total account value against a control.
When a campaign built this way still spends very little, the Performance Max not spending diagnosis checks targets, assets, feed, goals, exclusions and budget in order.
For the wider choice of which queries belong in Performance Max and which in a Search campaign, the comparison of Performance Max and Search campaigns sets out control, overlap and when to run each.
Common mistakes
- Splitting campaigns by category instead of by target. Many small campaigns, each starved of conversions, none learning well.
- Treating audience signals as targeting. Expecting the campaign to stay within a list it was only told to start from.
- No brand exclusion. Brand searches absorbed by Performance Max and reported as its wins.
- Final URL expansion left on with no exclusions. Clicks landing on the careers page.
- Judging the launch at campaign level. Search looks worse, Performance Max looks great, the account total is flat.
- Changing targets and assets weekly. Repeated learning periods and a campaign that never settles.
- One asset group with every product and every message. Assets that fit nothing well; no way to see which theme works.
- Ignoring the product report. A catalogue where a large share of products never receive an impression and nobody has asked why.
