How long Performance Max takes to settle

Performance Max is the campaign type people are least patient with, and impatience is precisely what breaks it. It is also the campaign type that most rewards leaving alone once it is set up properly.

How long does Performance Max take to stabilise?

Plan for two to six weeks. Where an account lands in that range depends almost entirely on conversion volume: a campaign generating 50 or more conversions a week will settle toward the short end, while one generating a handful can take well over a month and will stay volatile even then.

The often-quoted rule of thumb is that a bid strategy wants roughly 30 conversions in 30 days before its output means much. Below that, what you are reading week to week is mostly noise, and reacting to it makes things worse rather than better.

What resets the learning period?

More than people expect, which is why so many accounts live in permanent learning:

• Changing the bid strategy, or the target ROAS or CPA by a large step
• Significant budget changes — a rough working limit is 20% in one move
• Adding or removing asset groups, or materially changing the assets in one
• Changing conversion actions, or which ones count as primary
• Editing audience signals

Each reset costs days. Three edits in a fortnight means the campaign never leaves learning, and the performance you are judging is the performance of a campaign that has never been allowed to finish learning anything.

What should I do during the learning period?

Very little, deliberately. Set it up properly, then leave it: no target changes, no budget swings beyond 20%, no asset-group surgery.

There is useful work to do that does not touch the campaign. Watch the search terms and the brand-versus-non-brand split. Check the product feed for disapprovals. Verify conversion tracking is accurate, because everything the algorithm learns comes from that data. Note what you would change, and change it after the campaign has settled rather than during.

What actually counts as broken

Waiting is the right default, but it is not unconditional. Intervene when spend is concentrating almost entirely on brand traffic you were already winning — the most common way Performance Max flatters its own numbers, and the reason brand exclusions matter. Intervene if the budget is not spending at all after a week, which usually means the feed or the targets, not the algorithm. Intervene if conversions are being recorded that did not happen, because you are training on fiction.

Those are structural faults. “ROAS was lower this week than last” is not one of them, and treating it as one is how accounts end up permanently reset.

The setup decisions that matter more than patience

Most Performance Max outcomes are decided before launch. Brand exclusions so the campaign is not billing you for traffic you already had. Asset groups split by product group rather than one group covering the whole catalogue. Audience signals that are genuinely informative — Customer Match lists and real converter data, not a broad interest category. A product feed with correct titles and attributes, because in Shopping-heavy accounts the feed is the targeting.

Get those right and the learning period is something you wait out. Get them wrong and no amount of patience fixes it. More on how I structure this on the performance marketing page.