Performance Max Campaigns: Optimisation Strategies Beyond Google’s Black Box

Google’s Performance Max promises to find conversions across every Google surface, Search, Shopping, YouTube, Display, Discover, Gmail, and Maps, through a single campaign type. The trade-off is opacity. You hand over targeting decisions, placement controls, and creative rotation to an algorithm that returns aggregated metrics with limited visibility into what’s actually happening. Most advertisers respond by either trusting the black box completely or abandoning PMax for traditional campaigns. Both are wrong. Performance Max can be optimised, not by overriding the algorithm but by feeding it sharper inputs, sharper structure, and sharper conversion signals. By the end of this article, you’ll know which levers actually move PMax performance and which ones waste your time.

The first mistake advertisers make is treating audience signals as targeting. They’re not, they’re seeds. Audience signals tell Google’s algorithm where to start looking, after which it expands aggressively into lookalike behaviour across every Google surface. If you upload your customer list as the only signal, PMax will explore far beyond it. This is by design. Your job isn’t to constrain PMax; it’s to give it directional clarity. The most useful signals are first-party customer lists segmented by purchase value, in-market and custom intent audiences mapped to your category, and demographic ranges that mirror your highest-LTV cohort. Treating signals as filters leads to under-spending and missed opportunity; treating them as direction unlocks scale.

A single PMax campaign can hold multiple asset groups, and how you structure them shapes performance more than any bid adjustment. The rule: one asset group should serve one creative job. For e-commerce, that often means splitting by product category, margin tier, or seasonal intent. For lead gen, it means splitting by service line or buyer persona. Mixing high-margin and low-margin products in the same asset group dilutes ROAS targeting because the algorithm averages performance across mismatched economics. Each asset group should contain at least five headlines, four descriptions, fifteen images, and one video. Google will fill creative gaps with auto-generated assets if you don’t, and auto-generated assets typically underperform your own.

PMax will eat your brand search if you let it. Without account-level brand keyword exclusions, your PMax campaigns absorb branded queries at low CPA and inflate reported performance, while your dedicated brand Search campaigns starve. Request brand exclusions through your Google rep or use the account-level negative keyword list available in most accounts. Search Themes, the relatively newer steering feature, lets you nudge PMax toward specific query categories without forcing exact-match targeting. Use them for new product launches or seasonal pushes where you want the algorithm to lean into specific intent buckets. Layer in a comprehensive negative keyword list at the account level for irrelevant categories, competitor terms you don’t want to bid on, and informational queries that don’t convert.

PMax performs only as well as the conversion data feeding it. For e-commerce, Max Conversion Value with a target ROAS aligned to your blended margin works better than Max Conversions with target CPA because it lets the algorithm chase higher-value purchases. For lead gen, you need a value hierarchy, a demo request shouldn’t carry the same weight as a newsletter signup. Assign conversion values manually inside Google Ads or pass dynamic values through enhanced conversions. New customer acquisition rules, which let you bid more for first-time buyers, are particularly useful for brands with strong repeat purchase economics. Without a value hierarchy, PMax optimises toward the easiest conversions, which are rarely the most profitable.

PMax reporting has improved but remains thinner than traditional campaigns. The Insights tab now exposes search categories driving conversions, top-performing audience signals, and asset performance scores. Pair this with publicly available PMax scripts, Mike Rhodes and Optmyzr both publish maintained versions that extract search terms, placement data, and asset-level conversion breakdowns into spreadsheets. Cross-reference with GA4 to validate channel-level revenue contribution against what Google reports. Splitting your account into multiple PMax campaigns by product category or geography also improves visibility because each campaign reports performance separately. The goal isn’t to micromanage PMax, it’s to know enough to make budget allocation decisions across campaigns and asset groups.

Performance Max rewards advertisers who give it structured inputs, accurate conversion values, and sharp signals — not those who fight the automation. Build asset groups around economic logic, exclude brand traffic at the account level, assign meaningful conversion values across your funnel, and use scripts and the Insights tab to recover the visibility Google removed by default. The advertisers extracting durable ROAS from PMax aren’t trying to crack the black box; they’re feeding it cleaner data than their competitors are.