Google Performance Max Setup: The Build, Step by Step

Performance Max is Google’s own definition of hands-off: a goal-based campaign type that serves across YouTube, Display, Search, Discover, Gmail and Maps from a single campaign. You hand it a goal, a budget, a pile of creative and some hints about your customers, and it decides the rest.

I’ve already written about the guardrails I set before any Performance Max campaign spends money - brand exclusions, negative keywords, placement exclusions, the protective side. This is the other half: the build itself, every screen where the setup flow asks you for a decision, and what I actually choose on each one. The two posts are a pair on purpose. The build decides what the machine chases. The guardrails decide what it may not touch.

Before the interface: two decisions on paper

The build goes wrong most often before anyone opens Google Ads, so I settle two things first.

What counts as a conversion, exactly. Performance Max optimizes toward the conversion goals attached to the campaign with total literalness. If a newsletter signup and a purchase both count, it will buy the cheaper one at volume. My rule is one primary action per campaign, valued honestly, with everything else set to observation - the same conversion hygiene that every Smart Bidding strategy depends on, because Performance Max is Smart Bidding plus creative and placement automation on top.

Whether Performance Max is even the right tool. Google positions it as a complement to keyword-based Search campaigns, not a replacement, and I agree with the vendor for once. If you need to control exactly who sees the ad or exactly which query triggers it, a structured campaign still does that better. Performance Max earns its place where you have solid conversion volume, real creative variety, and goals you can express as a number.

Step 1: objective and bidding

The flow opens by asking for an objective and a bid strategy, and this is the one screen where I’d slow down. Your realistic options are Maximize conversions, optionally with a target CPA, or Maximize conversion value, optionally with a target ROAS.

My default for a new campaign is the plain Maximize option with no target for the first few weeks. A target set before the campaign has data is a guess wearing a settings toggle, and a wrong guess throttles delivery while everyone wonders why nothing serves. Once there’s a few weeks of stable conversion data, I add the target the numbers support. The full decision logic, including how much conversion volume each strategy wants, is in the Smart Bidding guide - it applies here unchanged.

Value-based bidding deserves one honest caveat: it only works if your conversion values mean something. An ecommerce feed with real order values, yes. A lead gen account where every form fill is recorded as $1, no - fix the values first or stay on Maximize conversions.

Step 2: budget, sized for the learning period

Google’s guidance is blunt about patience: the campaign ramps up for one to two weeks and wants at least six weeks before you judge it. So the budget question is not “what can we afford today” but “what can we hold steady for six weeks without a crisis meeting.” I covered the organizational side of that promise in the guardrails post. At setup time, the practical rule is a daily budget you will not be forced to slash mid-learning, because every big change restarts the clock.

Step 3: the settings screen most people click past

Three things I always touch on the campaign settings screen. Locations, set to “presence” rather than the default that includes people merely interested in your locations - a quiet source of irrelevant spend for any business that ships or serves locally. Final URL expansion, which lets the machine pick landing pages beyond the ones you chose, and which I switch off or fence with exclusions at launch, for reasons the guardrails post walks through. And the automatically created assets toggles, which decide whether Google may generate ad text and video on your behalf. I leave those off for the first build too. I want week one to show me my own creative, not the machine’s remix of my homepage.

Step 4: asset groups, the actual work

The asset group is where the build stops being checkbox work, because this is the raw material the machine assembles into every ad on six surfaces. Google’s specs per asset group: 3 to 15 headlines, 1 to 5 long headlines, 2 to 5 descriptions, at least 4 horizontal and 4 square images with room for up to 20 of each, 1 to 5 logos, and up to 15 videos per orientation. The full spec sheet, and what asset quality actually buys you, is its own guide: Performance Max assets.

Two rules make the difference between an asset group and a pile of files.

One theme per asset group. I build one asset group per product line or audience story, not one giant group with everything. The assets inside a group get combined freely, so everything in it has to make sense alongside everything else - the same combination logic as responsive search ads, at larger scale. Distinct angles across headlines, not one message rephrased fifteen times, and the ad copy prompts earn their keep here, since a full asset group is a genuine volume of copywriting.

Bring your own video, even a modest one. Google’s page says it plainly: if you don’t add a video, one or more may be auto-generated from your other assets. Auto-generated video is exactly as good as you’d expect from an automated slideshow of your images. A simple honest video you made beats it, and if you truly have none, budget an hour with any modern editing tool before launch rather than accepting the machine’s collage on YouTube.

Step 5: audience signals are hints, not fences

The last build screen asks for audience signals, and the name is doing careful work. Per Google’s documentation, signals are “audience suggestions” that help the AI start learning, and Performance Max “may show ads to relevant audiences outside of your signals” when it predicts a conversion. Nothing you enter here restricts anything. It only accelerates the first weeks.

So I feed it the strongest data I have, in order: customer lists, site visitors who converted or nearly did, then custom segments built from the search terms my Search campaigns already convert on. What I don’t do is agonize over it. The signal shapes week one and fades. The same expansion logic runs Meta’s Advantage+ audience, and in both cases the real boundaries are the exclusions, not the suggestions.

Step 6: the guardrails pass, then launch

Before the campaign goes live, the full pre-launch guardrails checklist runs: brand exclusions so the campaign can’t claim your branded traffic, negative keywords from the queries your account already paid to regret, account-level placement exclusions, and the customer-acquisition setting if new customers are the goal. In my order of operations that pass is not optional polish. It’s the difference between a campaign whose numbers you can read and one that flatters itself with conversions you’d have gotten free.

How to read Google’s 27% claim

Google’s case for the format leans on one number: advertisers who adopt Performance Max see an average of 27% more conversions or conversion value at a similar CPA or ROAS, per Google’s own global data from late 2023. I read that the way I read Meta’s 20% claim for Advantage+ and every other number a platform publishes about its own product: the seller grading its own homework, averaged across accounts that may look nothing like yours, measured by the platform’s own attribution.

The claim is probably directionally true - the format is genuinely strong for accounts with clean conversion data and real creative. But “27% more conversions” as reported by the system that counts the conversions, before your brand exclusions were set, is not a forecast for your account. Your number arrives at the six-week mark, judged against your own margin math, and not before.

After launch: you can finally see the channels

For most of Performance Max’s life, its biggest legitimate criticism was that spend across six surfaces reported as one opaque blob. That changed with channel performance reporting, which breaks out impressions, clicks, conversions, conversion value and cost by channel - Search, Display, YouTube, Discover, Maps, Gmail and search partners - for any date range after June 6, 2025.

I treat that report as the campaign’s honesty check at every review. If most of your conversions come from Search and Shopping while Display absorbs the impressions, you’ve learned what the campaign actually is for your account, and that shapes whether the next dollar goes here or into a structured campaign. This audit-the-machine work, rather than bid-setting, is most of what the PPC job is now. And when the review turns up numbers you don’t like, the troubleshooting guide is the diagnostic order I run before changing anything.

FAQ

How many asset groups should a new Performance Max campaign have? As many as you have genuinely distinct themes, and no more. One product line or audience story per group, each meeting the minimums with real variety. Three focused groups beat eight thin ones, because a group without enough distinct creative gives the machine nothing to learn from.

Do I need a Merchant Center feed? Only for retail. If you sell products, linking Merchant Center turns Performance Max into your Shopping infrastructure and the feed becomes your most important asset group input. Lead gen and service businesses run the format without one.

Can I run Performance Max and Search campaigns together? Yes, and Google recommends exactly that pairing. Your keyword-based Search campaigns keep the queries you know convert, an exact-match keyword generally takes priority, and Performance Max hunts everywhere else. The brand exclusion from the guardrails list is what keeps the boundary honest.

How long before I judge the results? Six weeks, by Google’s own evaluation guidance, with the first one to two weeks of ramp excluded from the analysis. Judging earlier mostly measures the learning period, and reacting to it restarts the learning period. Check the guardrails early and often. Judge performance late.

This post is part of the paid media guides. Its other half is the Performance Max guardrails checklist, the bidding layer underneath is the Smart Bidding guide, and Meta’s version of the same delegation is the Advantage+ setup. The wider practice starts at the complete guide to AI marketing.