PMax is the only campaign type in Google Ads that shows ads simultaneously on Search, Shopping, YouTube, Gmail, Display, Discover, and Maps. One campaign instead of six.
You provide three things: budget, conversion goal, and ad assets. Google decides where, when, and to whom to show the ads.
Sounds convenient. But there is a downside: if the input data is incorrect, the algorithm spends the budget on irrelevant audiences. It is efficient, but not in your best interest.
After over 13 years in digital marketing, the picture is clear.
The more products in the feed, the more data for the algorithm. An online store with a catalog of thousands of parts is an ideal candidate. A store with only 10 items will not give PMax enough volume for learning.
Auto parts, dentistry, repair services – where the client searches for a specific product and is ready to buy.
Nor for businesses without analytics. We have seen accounts where PMax seemingly generated conversions – but in reality, Google counted a page view of the contact page as a target action. The owner thought everything was working. There were no real leads.
“Over years of practice, the statistics are unambiguous: PMax delivers the best CPL for e-commerce. But without a clean feed and correct signals, it will burn the budget just like any other campaign. Preparation is everything”
– Volodymyr Kashalaba, CEO Guild of Marketing
We covered the theory. Now – step-by-step.
Three things you should not launch PMax without.
Verify that real actions are tracked: purchase, lead form, phone call. Not page views. Not scrolling. In one of our projects, we discovered that three previous campaigns spent tens of thousands of dollars without a single recorded conversion. Conversions existed, but no one had set them up. Google simply didn’t know what to optimize for.
Current prices, real stock availability, proper photos, correct categories. A common problem: half the products in the feed are marked “in stock,” but they are not on the warehouse. Google spends the budget advertising products that cannot be bought.
Clinic, store, service center – link the profile to Google Ads. For one of our clients with a service center in Kyiv, this created a separate channel: clients searching for “pneumatic suspension repair nearby.”
Goal “Sales” for e-commerce, “Leads” for services. Type: Performance Max.
Budget: We recommend starting from $25-$35/day for e-commerce, from $15/day for services. With a lower budget, the algorithm will not gather enough data for learning. The learning period is 2-4 weeks. During this time, do not make changes.
Bidding strategy: Start with “Maximize conversions” without a target CPA (CPA – target cost per acquisition). Let Google determine the cost per conversion itself. After 30-50 conversions, switch to “Target CPA” with a CPA 10-20% higher than the actual one.
Asset Group (group of ad assets) – a set of headlines, descriptions, photos, and videos. Google assembles ads from them for each platform.
If the catalog covers several categories – create a separate Asset Group for each. Tires and rims in one group means a lack of segmentation that hurts conversion rates.
Headlines – up to 15, long (up to 90 characters) and short (up to 30). Real search queries, not marketing slogans. Example: “Mercedes air springs – from $125 with warranty”. Not “Quality auto parts at the best prices”.
Descriptions – up to 5, up to 90 characters. Specificity. Not “Wide assortment at affordable prices”. But “2,300 parts in stock. Delivery via Nova Poshta in 1-2 days”.
Images – minimum 5, preferably 15+. Horizontal (1200×628), square (1200×1200), vertical (960×1200). Product photo for Shopping, lifestyle for Display.
Video. Without your own, Google will stitch a slideshow. The result is unprofessional. Even a 15-second video shot on a smartphone will work better.
Signals are hints to the algorithm on who to target. This is not rigid targeting like on Facebook – Google can ignore them. But without signals, the start is longer and more expensive.
All this is setup. Now – what happened in practice.
Our client is an online auto parts store for premium cars. A narrow niche, 20+ brands. Average check from $125 to $750.
We did not rush into ads immediately. The first three months were for SEO foundation: GA4, Search Console, semantics, meta tags. PMax works better on a site with quality content and structure – the algorithm considers user behavior from all channels.
In parallel – WordPress updates, backups, security monitoring.
In parallel, we audited old campaigns. We found three Search campaigns that were burning the budget without a single conversion. The reason was simple: conversion goals were not set up. The algorithm optimized for clicks and delivered them reliably. There were clicks, but no sales.
Results are good. But PMax does not work for everyone. Here is where it burns most often.
We have seen each of these dozens of times during audits.
Google doesn’t know what a sale means for you and optimizes for clicks. There are clicks. No sales. Before PMax, import conversions from GA4 to Google Ads.
“Buy tires” and “buy rims” – different intents. Different headlines, different photos. For one client in the tire niche, we even separated summer and winter tires into different groups.
Outdated prices, missing photos, products marked “in stock” with zero inventory. The client clicks, sees “out of stock,” and leaves. You pay for a click that cannot become a sale.
The algorithm starts blind, testing everyone in a row. The first weeks are expensive and unstable. With signals – faster start and lower CPL at the beginning.
The first week CPL is high – the advertiser starts tweaking settings. Every change restarts the learning. A vicious cycle. Rule: Launch – do not touch for two weeks.
“In PMax, setup is 30% of the result. The remaining 70% is letting the algorithm work. Over 13+ years, the pattern is the same: those who give time get CPL many times lower. Those who fiddle daily just spin their wheels”
– Volodymyr Kashalaba, CEO Guild of Marketing
Separately, what PMax does not do on its own.
PMax does not kill other campaigns. It complements them. For one of our clients – a chain of tire stores – we run PMax, Search, Display, and remarketing simultaneously.
PMax takes the product catalog. The algorithm selects the product for each user and distributes impressions across platforms.
Search – for services and expensive queries with full control. In our auto parts project, the search campaign for repairs delivered conversions with a higher average check than the product PMax.
Remarketing brings back those who left to compare. In tires, a buyer visits 5-6 sites before deciding. In our project, Display remarketing delivered a comparable number of conversions to the main search campaign.
First 2-4 weeks – gather data. Then optimize.
In Insights, check what queries show your ads. Irrelevant ones – add as negative keywords. Google expanded negative keyword support for PMax: they are now available at both the campaign and asset group levels.
Google rates each element as Low, Good, or Best. Low is a candidate for replacement. Change 2-3 at a time, not all at once.
One category gives 80% of conversions, the other just eats the budget? Separate them into different Asset Groups or separate campaigns.
Increase budget by 15-20% per week, no more. A sudden doubling restarts learning – the algorithm is used to working with one data volume and needs time to recalculate patterns with a sharp change