Cluster 6 · Google Ads

Performance Max Best Practices (Without Losing Control)

By Syed AsifSeptember 30, 20267 min read

Good PMax structure: themed asset groups, audience signals per group, brand separated — one campaign serving all Google inventory.

In short

Performance Max works best when you give Google's automation the right structure instead of micromanaging it: themed asset groups (never one giant group), audience signals per group, brand terms excluded or separated, URL expansion controlled, and a clean product feed. Set the guardrails well, then let it learn.

Performance Max is the campaign type advertisers love to complain about: a black box that spends your budget across all of Google's inventory with minimal reporting. But in my Google Ads management work, PMax is also consistently one of the best performers — when it is structured properly. The advertisers losing control are usually the ones who gave the algorithm nothing to work with: one asset group, mixed intents, no exclusions, dirty feed. This article is my setup playbook.

What PMax actually does (and doesn't)

Performance Max is a goal-based campaign that serves your assets across Search, YouTube, Display, Discover, Gmail and Maps from a single budget, optimizing toward your conversion goal with Smart Bidding. What it does well: find incremental conversions you would never have targeted manually, especially in Shopping and remarketing.

What it does not do: respect your brand strategy by default (it will happily spend on your own brand terms and take credit), separate new from existing customers without configuration, or tell you where anything happened. Control in PMax does not come from bid adjustments — it comes from structure, assets and exclusions. That is the entire philosophy of this playbook.

Setup best practices

Separate brand from non-brand

The single highest-impact PMax decision. Create brand exclusions on your prospecting PMax campaign and run brand demand in its own campaign. Otherwise PMax harvests cheap brand conversions, reports a beautiful ROAS, and you learn nothing about incremental performance.

One asset group per theme

Asset groups are PMax's version of ad groups. Each should represent one product line, service or audience theme with its own tailored headlines, images and audience signal. Two to five asset groups cover most advertisers. The algorithm learns per asset group — mixing everything into one teaches it nothing.

Audience signals, not audiences

Audience signals are starting points, not targeting. I feed each asset group: custom segments (competitor sites, relevant search terms), first-party lists (site visitors, customer uploads), and in-market segments that match the theme. Good signals shorten learning dramatically.

Control URL expansion

URL expansion lets Google send traffic to pages beyond your final URL. I keep it on only when the site's page quality is uniformly high; otherwise it finds your weakest pages. For lead-gen, I usually turn it off and point each asset group at its dedicated landing page.

Feed hygiene (Shopping advertisers)

For ecommerce, the product feed is the campaign. Titles optimized for search terms, correct GTINs, accurate availability and pricing, strong images. I have seen feed cleanup outperform every other PMax optimization combined.

Setting → recommendation

Setting	My recommendation	Why
Brand exclusions	Exclude brand from prospecting PMax	Stops PMax taking credit for existing demand
Asset groups	2–5 themed groups, never one	Lets the algorithm learn per intent
Audience signals	Custom + first-party per group	Shortens learning, guides prospecting
URL expansion	Off for lead-gen; on for strong ecommerce sites	Prevents traffic to weak pages
New customer acquisition	Set higher value for new customers	Stops paying full price for existing customers
Account-level negatives	Add irrelevant/brand terms	The only negative control PMax offers
Budget	Enough for 30+ conversions/month per campaign	Below this, learning never stabilizes

Asset quality matters more than you think

With Smart Bidding doing the targeting, your assets are your main lever. I fill every slot: 15 headlines, 5 descriptions, all image sizes including vertical, and a real video (even a simple one — Google auto-generates video from images otherwise, and it looks it). Headlines should cover the theme's angles: offer, proof, urgency, feature, brand. Weak assets are the quiet reason most PMax campaigns underperform — the algorithm can only assemble combinations from what you give it.

Measurement: the unsexy part that decides everything

PMax optimizes toward whatever you tell it a conversion is worth. If your conversion tracking is dirty — double-counting, inflated values, missing offline conversions — PMax will brilliantly optimize toward the wrong thing. Before launch I verify: enhanced conversions on, values assigned honestly, new-vs-existing customer value rules set, and (for lead-gen) CRM import of qualified leads where possible. Pair this with the relevance discipline from my Quality Score work: strong landing pages lift every campaign type.

Common PMax mistakes

One giant asset group. Mixed intents, meaningless learning, useless reporting.
No brand exclusion. Beautiful ROAS, zero incremental growth.
Editing during learning. Touching structure or budgets in the first 2–3 weeks restarts learning. Set it right, then hands off.
Dirty feed. For Shopping, a neglected feed caps everything.
Judging on 7 days of data. PMax needs 2–4 weeks minimum. Impatience is the most expensive PMax setting.
No Search campaign alongside. PMax plus a tightly controlled Search campaign for core intent keywords beats PMax alone — they cover each other's blind spots.

Stop trying to outsmart the algorithm and start giving it better raw material: clean structure, strong assets, honest conversion data. PMax rewards good inputs more than clever management.

If your PMax campaigns feel like a black box eating budget, I audit them against this exact checklist — structure, exclusions, assets, feed and tracking. Get an honest read on what's actually happening in your account.

Frequently asked questions

Neither is universally better. Search gives you intent control and transparency; Performance Max gives you reach across all Google inventory with automation. I run Search for high-intent keywords I want to control precisely, and PMax for incremental reach, remarketing, and Shopping — with brand properly excluded or separated.

Use campaign-level brand exclusions to block your brand terms from a non-brand PMax campaign, and run a separate brand campaign (PMax or Search) to capture that demand cleanly. Also add brand terms as account-level negative keywords where appropriate. This stops PMax from taking credit for conversions you would have gotten anyway.

One asset group per distinct product line, service, or audience theme — typically 2 to 5 for most advertisers. One giant asset group mixing everything is the most common PMax mistake: the algorithm cannot learn which assets serve which intent, and reporting becomes meaningless.

It can, but with limits. PMax needs conversion volume to learn — I recommend at least 30 conversions per month per campaign for stable learning. Below that, keep the structure simple (one campaign, tight asset groups) and use Maximize Conversion Value only once you have reliable value tracking.

Expect 2 to 4 weeks of learning before judging performance, longer on small budgets. Do not touch structure, budgets, or assets during learning unless something is clearly broken. Most PMax "failures" I am asked to fix are campaigns that were edited to death in week one.

Syed Asif

SEO Consultant & Digital Marketing Freelancer with 5 years of experience and 150+ projects. I help businesses rank on Google and get cited by AI search.

About Syed Asif →

PMax feeling like a black box?

I audit Performance Max accounts against this checklist — structure, brand exclusions, assets, feed and conversion tracking. Honest read, no obligation.

Get My Free Audit WhatsApp Me