Performance Max Guardrails to Set Before Going Live

Performance Max is the most automated campaign type Google sells: one campaign, every Google surface, with the machine choosing queries, placements, audiences and creative combinations on your behalf. That design has a consequence the onboarding flow does not dwell on. Every control you skip at setup is a decision you have delegated to a system whose objective is conversions as you have defined them, on inventory it profits from selling you. Some of those delegations are fine. A few are expensive, and all of the expensive ones are cheap to prevent before launch and annoying to unwind after.

So before any Performance Max campaign of mine goes live, the guardrails below are in place. None of them requires a developer, all of them live in the Google Ads interface, and together they take under an hour - against a campaign that will spend for months. This is the protective half of the job. The constructive half - bidding, budget, asset groups, audience signals, in order - is in the setup walkthrough, and the two are meant to be read as a pair.

Brand exclusions, so the machine can’t claim your brand

Left alone, Performance Max will serve on searches for your own brand name, and it will report those conversions as campaign performance. Those are the cheapest, most certain conversions your account will ever record - people who searched for you by name - and a campaign that absorbs them looks brilliant while adding little. This is the first guardrail I set, because it protects the integrity of every number that follows.

The tool is brand exclusions, applied at the campaign level. In Performance Max they cover both Search and Shopping ads, and Google’s documentation is direct about why they beat a do-it-yourself negative keyword list for this job: an excluded brand blocks misspellings, foreign-script variants and related sub-brands automatically, which a keyword list catches only as fast as you type them in.

If part of your strategy is a separate branded Search campaign - usually the right call - the exclusion also stops Performance Max cannibalizing it, and your brand traffic stays where you can read it cleanly.

Negative keywords, for the queries you already know are bad

Performance Max now accepts campaign-level negative keywords, which closed the loudest control gap in the product’s history. Two things to know about their reach. They apply to Search and Shopping inventory only, so they do not filter video or display placements. And account-level negative keywords apply to Performance Max automatically, which makes the account list the right home for the permanent exclusions - competitors you never bid on, job-seeker terms, free-seeker terms, support queries from existing customers.

My starting list for a new Performance Max campaign is not speculative. I pull it from the search terms the account’s existing Search campaigns already paid for and regretted: every query theme that spent money without converting goes in before launch. The machine will find plenty of new queries. It does not need to re-learn the failures I have already paid to discover once.

Placement exclusions, for where your ads should never appear

Outside Search and Shopping, Performance Max places ads across YouTube, display inventory and apps, and the campaign settings offer little control over where. The control exists one level up: Performance Max respects account-level placement exclusions, where you can exclude specific sites, YouTube channels and app categories for every campaign at once - up to 65,000 entries per account.

Before launch I exclude the standing categories that almost no advertiser wants: made-for-kids app inventory, parked domains, and any content themes the brand cannot sit beside, which is also where the brand suitability settings for inventory type earn a review. This is a fifteen-minute job the first time and near zero afterwards, because the account-level list carries over to every future campaign.

One honest caveat: exclusion lists are a coarse instrument, and you will not anticipate everything. That is what the placement report in the first-week checks is for.

The settings that steer spend: URL expansion and customer type

Two switches in campaign setup decide more than their screen real estate suggests, and I check both deliberately rather than accepting the defaults.

Final URL expansion. Switched on, Performance Max may send clicks to any page on your site it judges relevant, with its own generated ad text, instead of only the landing pages you chose. For a tightly managed funnel I switch it off at launch, or keep it on with URL exclusions covering careers pages, legal pages, the blog and anything else that should never be a paid landing page. Expansion can be a useful query-discovery tool later. In week one I want to know exactly which pages my spend lands on.

Customer acquisition. If the campaign exists to acquire new customers, the “new customers only” setting - fed by your customer lists - stops the machine claiming conversions from people who already buy from you. Without it, retention conversions inflate the campaign the same way brand traffic does. With it, the reported number is closer to the thing you actually wanted bought.

Conversion goals: the guardrail inside the guardrail

Every guardrail above limits where the machine spends. This one governs what it chases, and it is the one I have seen do the most damage when skipped. Performance Max optimizes toward the conversion goals attached to the campaign, exactly as configured. If newsletter sign-ups, PDF downloads and purchases all count as conversions with equal weight, the machine will cheerfully buy you the cheapest of the three at volume and report success.

Before launch: the campaign’s goals include only the actions that constitute success for this campaign, secondary actions are removed or marked as observation-only, and values reflect reality if value-based bidding is in play. This is the same discipline as anywhere in Smart Bidding - the bidder optimizes exactly what you count - but Performance Max raises the stakes, because the same goals steer creative and placement choices on top of the bids.

Budget, ramp, and agreeing not to panic

Guardrails also apply to the humans. Google’s own guidance for Performance Max is explicit that the campaign ramps up for one to two weeks after launch and needs at least six weeks before evaluation, with the ramp excluded from analysis. So the last pre-launch step is a calendar agreement, made while everyone is calm: the evaluation date, the metrics that will count, and a commitment not to restructure the campaign at day ten because the early numbers wobble. Every panicked edit restarts learning, and I have watched more Performance Max budgets die of thrashing than of bad targeting.

I set the opening budget at a level the business can tolerate for that full learning window without anyone demanding intervention. A number that forces a week-three crisis meeting was the wrong number, whatever the forecast said.

The first-week checks

Guardrails before launch, verification after. Three things I read in the first days of any new Performance Max campaign:

1. The placement report. Performance Max has a dedicated placement report showing where ads actually served. It is impression-level rather than cost-level, which is a real limitation - but it is enough to spot placement categories worth adding to the account exclusion list while the numbers are still small.
2. Search themes versus reality. The insights panel shows the query themes the campaign is spending against. Anything embarrassing goes into negatives now, not at the month-end review.
3. Where conversions are coming from. If early conversions cluster suspiciously - all from brand-adjacent queries, all from existing customers - one of the guardrails above is misconfigured, and week one is when that costs the least. The arithmetic of what a conversion is genuinely worth belongs to the ROI method, and it is much easier to defend when the campaign’s numbers were clean from day one.

The theme across all of this: Performance Max is not a campaign you set live and inspect quarterly. It is a delegation, and the guardrails are the contract’s terms. Written before launch, they are configuration. Written after, they are cleanup. And if the campaign is already live and disappointing, the troubleshooting guide picks up exactly where these checks leave off. This constraint-writing work is, increasingly, what the PPC manager’s job actually is now.

This post is part of the paid media guides, where the same spend discipline runs through choosing a Smart Bidding strategy, Meta’s version of this delegation and evaluating ads inside ChatGPT. The retroactive version of these fences, for accounts already running, is the wasted spend audit. The creative the machine assembles still has to be written - the ad copy prompts cover that - and the wider practice starts at the complete guide to AI marketing.