Performance Max for Small Budgets: Make Google's AI Ads Behave

What Performance Max actually does with your money

Performance Max is a single campaign type that serves ads across Search, Shopping, YouTube, Display, Gmail, Discover and Maps. You supply assets, a budget and conversion goals; Google's systems decide where, when and to whom your ads appear. There are no keywords to bid on and, historically, very little visibility into placements, though reporting has improved and you can now see more about how spend splits across channels.

That trade works well when the algorithm has plenty of conversion data to learn from. On a small budget it often does not, and an under-fed PMax campaign tends to chase volume on the cheapest available inventory, which is usually low-intent Display and YouTube space. The result looks fine on click counts and dreadful on enquiries. Everything in this guide is about preventing that failure mode.

When you should stick to Search instead

Under £1,000 a month, a standard Search campaign with exact and phrase match keywords is often the better tool, because every click comes from someone actively searching for what you sell.

• Choose Search if you are lead generation with a tight service area and modest conversion volume
• Choose Search if you have no product feed; PMax without Shopping data loses much of its edge
• Consider PMax if you run ecommerce with a well-maintained Google Merchant Center feed
• Consider a feed-only PMax (no other assets) as a low-risk way to get Shopping coverage
• Run brand searches in their own Search campaign regardless, so you control that traffic

A sensible small-account pattern is Search for your core services, plus feed-only PMax if you sell products online. Full PMax with all asset types is a later step, not a starting point.

Asset group structure that concentrates learning

Asset groups are PMax's closest equivalent to ad groups. On small budgets, restraint wins: two or three asset groups, each built around one clear product or service theme, so conversion data concentrates instead of fragmenting.

• One theme per asset group, with final URLs pointing to the matching landing page, never the homepage
• Fill every asset slot: headlines, long headlines, descriptions, images in all ratios and your logo
• Supply your own video; if you do not, Google auto-generates one from your images, and the results are rarely on-brand
• Use search themes (up to 25 per asset group) to tell the system which queries matter
• Add audience signals from your customer lists and site visitors; they steer early learning even though Google treats them as suggestions

Negative keywords, brand exclusions and the other guardrails

PMax launched with almost no levers, but Google has gradually relented, and the current controls are enough to keep a small budget honest if you actually use them.

• Negative keywords: maintain an account-level negative list for irrelevant terms, and use the campaign-level negative keyword support Google has added to PMax for campaign-specific exclusions
• Brand exclusions: stop PMax hoovering up your own brand searches and claiming them as wins; serve those through your dedicated brand Search campaign
• Final URL expansion: turn it off, or add URL exclusions, so Google cannot send traffic to your careers page or blog
• Content suitability and placement exclusions: review the placement report regularly and exclude junk apps and sites
• Location settings: check you are targeting presence, not interest, or a West Bromwich plumber will pay for clicks from anywhere

Feed the algorithm honest conversion data

PMax optimises to whatever you count as a conversion, so counting the wrong thing guarantees the wrong outcome. Page views, all-calls-of-any-length and newsletter sign-ups should never be primary conversion actions on a small account.

Set genuine enquiries, qualified calls and purchases as primary; demote everything else to secondary. For lead generation, assign proxy values reflecting what different enquiry types are worth, and if your sales cycle allows it, import offline conversions so the system learns which leads became customers. Enable enhanced conversions so measurement survives the consent-banner era. Clean data does more for a small PMax campaign than any bid tweak.

A 90-day plan and when to pull the plug

Weeks one and two: build the structure above, verify conversion tracking with test enquiries, and set a daily budget you can sustain without flinching. Weeks three to eight: leave it alone. PMax learning genuinely takes weeks, and every structural edit partially resets it; limit yourself to adding negatives and excluding junk placements.

From week eight, judge it on cost per conversion against your Search baseline, not on clicks or impressions. Review search themes, placements and channel splits monthly. If after ninety days PMax still costs materially more per enquiry than Search, move the budget back without sentiment; the tactic will still be there when your conversion volume grows. If you would like an experienced pair of eyes on the setup, our team at Thind Global Services can help.