PMax Campaign Setup: What to Check Before You Build It

This is how we set up new non-brand Performance Max campaigns. The method doesn't depend on any particular tool: every step works by hand in the Google Ads UI and in Merchant Center, and an AI assistant takes over the lookups and the math. The video above shows that work in a live store's account: the assistant reads the feed and the account, asks about scope, and ends with a preview of the changes. We masked the industry, names, and amounts. A Polish audio track and Polish subtitles are available in the player settings.

What the AI handles and what you decide

Most of the time on a new campaign goes into lookups: how many products in the category are in the feed and how many are eligible to serve, where they sell today, which negative keyword lists and brand list are attached to the other campaigns, what the location settings are, and which conversion goals apply. An AI assistant with access to the account and to Merchant Center gets through all of that in a few minutes. It:

• reads the feed and matches the category's products against their results from the last 90 days,
• checks which campaigns those products spend money in today,
• pulls location settings, negative keyword lists, and the brand list from the other campaigns,
• flags products that get impressions and zero clicks,
• builds a preview of the exact changes that would be sent to the account.

The decisions stay with you: which products go in, what happens to products that already sell elsewhere, and whether and when to turn the campaign on. Nothing reaches the account without a yes, and the campaign is created paused.

Step 1: Check whether a separate campaign makes sense

Every campaign learns from its own conversions. Ten campaigns with one conversion a month each do worse than one campaign with ten, so the first question is how many conversions a month the category brings in today.

• Bidding to a target ROAS needs at least 50 conversions a month per campaign, and it works well at 60 to 90.
• With a handful of conversions a month, a separate campaign has nothing to learn from. You're better off leaving the category where it is and excluding the products you don't want.
• A store with no converting Shopping campaigns should start with a Standard Shopping campaign on manual CPC and add Performance Max once that campaign has collected about 15 conversions.
• Products with different margins shouldn't share a campaign, because the campaign optimizes toward one ROAS for everything in it. With mixed margins, a single target ends up pushing products the store loses money on.
• An account can have at most 100 Performance Max campaigns, enabled and paused combined. With a large catalog, you hit that limit sooner than you'd expect.

You also need your break-even ROAS, which is 1 ÷ margin. At a 25% margin, break-even is 400%, and at a 40% margin it's 250%. Without the margin, any target ROAS is a guess.

Step 2: Start with the feed

The feed sets the scope of the campaign, so it comes before any talk of budget. The first table to pull shows how many products the category has, how many are eligible, and what clicks, conversions, and return they brought in over the last 90 days. That table tells you whether the category has sales history, which is what you'd base a target on.

• Take product counts, availability, and disapproval reasons from Merchant Center. The product report in Google Ads shows a narrower set and leaves out why products were disapproved, so confirm the account's numbers in Merchant Center before you settle the scope.
• "Not eligible" doesn't necessarily mean the product is disapproved in Merchant Center. The most common reason is "no campaign is advertising this product," or an exclusion in a campaign's product tree, which is a targeting decision. Read the reason before you treat a product as broken.
• Check whether the feed carries the store's own categories (product_type) and custom labels. An automatically generated feed, which Merchant Center builds by crawling the website, has neither, so you can't split the catalog by store category. You're left with brand, or with adding a label through a supplemental feed, a second data source that fills in attributes of existing products. A crawled feed as the only product source is a weak base for campaigns in any case; connect a feed from the store's platform if you can.
• Decide on the scope: the whole category or only some subcategories.
• Keep new products with no conversion history out of a campaign that makes money. Test them separately and move them over once they make their first sale.
• Limit products on the campaign side or with labels. Removing them from the feed takes data away from other campaigns and from free listings.

Step 3: Find out where these products sell today

A quick setup in the UI usually skips this step, yet this is the step that decides whether the new campaign adds sales or just cannibalizes other campaigns. Line up the category's products against every campaign they spent money in over the last 90 days. Often these are campaigns that already split the catalog by performance, such as top sellers and everything else.

The same products shouldn't overlap between two Performance Max campaigns. Two campaigns on the same products don't drive up each other's bids, because one account doesn't bid against itself. What they do is split the demand and the data the bidding learns from. Each sees only part of the conversions, and neither holds a consistent target. For the same reason, two campaigns on the same products give you a contest between them, which is useless as a test. That leaves three choices:

• exclude the category from the existing campaigns and move it to the new one,
• leave the category where it is and skip the new campaign,
• move only some of the products and leave the rest in place.

Moving products comes with a counterintuitive rule. The older campaign with history keeps the best products, and its budget is set about 20% above what those products used to spend in it. The rest goes to the new campaign. Pulling your bestsellers into a new campaign usually costs a few days of sales while the new campaign ramps up. If the category needs its own budget or target and you move it whole, bestsellers included, plan for that dip and judge the category as a whole. Any product you move into the new campaign gets excluded from the campaigns that had it before. That applies to every Performance Max campaign carrying those products, including small ones that spend little on them.

Order matters. Build the new campaign with the products included in its tree, confirm that it serves, and only then exclude the products from the older campaigns. For 12 to 24 hours the products serve in both places, so work out that extra spend before you turn anything on. Doing it the other way around leaves the products without ads for days.

Moving a product doesn't wipe its history. Google learns about products across the whole account, so a new campaign with products that already sold ramps up faster than one with products that never sold. The first days after moving bestsellers can still be weaker, because the campaign learns its own return from scratch.

Two notes on measuring overlap. Count it only across enabled campaigns and only for products that actually spent money, and leave the brand campaign out. Without those filters, the overlap can come out dozens of times higher than it really is. A feed-only campaign serves mostly in Google Shopping, so it overlaps little with Search campaigns, though some of its budget can end up in Search and video. Cost by channel and a comparison of both campaigns' search terms show how much.

Step 4: Brand stays in the brand campaign

Conversions from searches for the store's own name would have happened without a Shopping ad, and in a non-brand campaign they inflate ROAS and throw off the target. The default setup excludes the brand from non-brand campaigns on day one and lets a separate brand campaign handle brand traffic, with a budget of up to about 10% of total account spend.

• In Performance Max you exclude the brand with a brand list. It matches searches to the brand as an entity in Google's database, so it misses typos, sub-brands, and part of the brand searches. Back it up with negative keywords in phrase match. The brand name alone in exact match lets "brand + product" searches through, and those can make up a few percent of a non-brand campaign's cost and inflate its ROAS.
• The store's brand may be missing from Google's database, or the first suggestion may be someone else's company. Search by the full domain and check that the website URL matches.
• A brand list rejects an entry when the brand name doesn't appear on the landing page. Before you decide the list is broken, check that the name is on the page.
• Brand typos and variants stay in the brand campaign. When you're unsure whether a search is a brand search, leave it alone.
• Keep the brand exclusion off the account level, because there it blocks the brand campaign too. A list attached to the non-brand campaigns is enough.
• There are exceptions. For a manufacturer, the brand name is often the category name, and for a very well-known brand, an exclusion hands the top spots to competitors. In that case the brand stays in the non-brand campaign, you steer it with a higher target ROAS, and you report it separately.

That's how we build accounts: the brand excluded from non-brand campaigns, plus a separate brand campaign. Whether an exception applies depends on how well known the brand is and whether competitors bid on its name.

Once the brand is cut out, the non-brand campaign's reported ROAS drops. The drop only means its number stopped counting traffic that would have come anyway.

Step 5: Attach the negative keyword lists right away

A new campaign inherits nothing from the others, including negative keyword lists and the brand list. In one account we manage, a new non-brand Performance Max campaign picked up brand searches for 16 days, and they made up 7.7% of its cost. Its reported ROAS of 876% was about 720% without them.

• Right after you create the campaign, check which lists are attached to the existing non-brand campaigns and attach the same ones.
• Check how many entries each list has. A list can be attached and empty.
• Attaching a list applies every keyword on it, and any change to the list takes effect immediately in every campaign it's attached to.
• One list holds up to 5,000 keywords, and a Performance Max campaign accepts up to 10,000 negative keywords.
• Account-level negatives block keywords in every campaign, including the brand campaign. Use negative keyword lists when the exclusions should apply only to specific campaigns.
• Placement and app exclusions for Performance Max exist only at the account level. They apply to every campaign in the account, and Google doesn't always enforce them fully. This mostly matters for campaigns with assets.

Before launch, it helps to sort candidate negatives into a four-column spreadsheet: the brand and its variants, words that bring bad traffic on their own, obvious junk, and searches to revisit after launch instead of excluding up front, such as competitor names.

Step 6: Set location to presence and check it against the source campaign

Set the campaign to reach people who are in or regularly in your locations. "Presence or interest" lets in traffic from outside the country you sell to, most of all in campaigns with assets.

Location settings are usually copied from the store's other campaigns, and copying carries over their mistakes too. In one account a new campaign picked up an extra region; in another, three cities that the source campaign had as exclusions ended up as targeting. So compare the new campaign's location settings with the source campaign's twice, once for targeted locations and once for excluded ones. After the campaign is created, check that it isn't set to "All countries and territories."

When the account covers several countries or Merchant Center has several data sources, specify the store and the country you're building the campaign for. A campaign without a feed label pulls products from every data source in the account. The language setting refers to the language of the user's interface and says nothing about the language of the search. In a non-English market, you can add English alongside the local language, and with presence targeting the campaign still stays in the target country.

Step 7: Check the conversion goals after you create the campaign

A new campaign starts with the account-default conversion goals, even when the other campaigns use campaign-specific goals. In one account the account-default goals included product views and checkout starts, with no purchase. For 13 days the new campaign bid toward product views. The conversions column showed zero, even though the campaign drove more than a hundred purchases in that time.

• After you create the campaign, check its goals and compare them with a campaign that performs well.
• For an online store, purchase should be the only primary goal. Using add to cart as a primary goal is a short-term workaround for when purchase volume is very low.
• Track purchases with the Google Ads tag. An import from Google Analytics 4 usually captures less data, and the campaign learns from what it gets.
• A custom goal doesn't switch off the account-default goals; it adds to them. The campaign then counts value twice, and your target ROAS no longer measures what it used to.
• Changing conversion goals changes what the campaign bids toward, so make it a deliberate decision of its own.

Step 8: Budget and target ROAS at launch

When the products have history in the account, their actual spend sets the budget. Base the new campaign's budget on what those products spent over the last 30 days. The budgets of the campaigns you're moving them from tell you nothing here.

• Once the category is excluded, the old campaigns keep their budgets. If you leave those alone, the new campaign's budget is new money, and the old campaigns spend the freed-up amount on their remaining products. Before you approve the change, call it what it is: a shift with the account total unchanged, or an increase, with the total before and after.
• For products with no history in an account that has history, start at about 30% of the average daily spend of a typical campaign in the account. In a new account or a new market, start at 10% of the maximum budget you work out from Keyword Planner data (searches × CTR × CPC). Whether a test means anything depends on how much of the search demand it covers: at coverage of a few tens of percent the result is repeatable, and below one percent it's noise.
• Google can spend up to twice the daily budget on any given day, but over a month it charges no more than 30.4 times the average daily budget.
• The first days often cost more, since a campaign without history bids harder to land its first conversions. A new campaign won't lower your cost per click.
• Check where launch and the window you base the target on fall in the category's season. Ahead of a peak, set the campaign up early, since it needs time to learn; in a slow season, launch with a lower budget. A window that includes a sale, Black Friday for example, inflates the return, and a target set on that number will choke the campaign once the sale ends.

Set the daily cap close to what the products naturally spend. A campaign that runs far below its cap buys weaker traffic: in one account, a campaign at 35% budget utilization had a 1.41% Conv.Rate, while a comparable campaign at 103% of budget had a 2.09% Conv.Rate at almost the same cost per click. When a campaign uses less than half its cap and its Conv.Rate trails a comparable campaign, lower the cap.

A feed-only Performance Max campaign starts with Maximize conversion value and no target ROAS. You add a target once the campaign reaches at least 50 conversions, at the level the campaign hits by then or 10 to 20% lower, and no higher. The exception is splitting a catalog into campaigns by performance, when the category has sales history and brings in at least 50 conversions a month. Then we set a target from day one, derived from the account target: the top-seller campaign gets 0.9 times the account target, the core group the full target, and costly products 1.2 times. When you work out the category's return, leave the brand campaign out, because those sales won't follow the products into the new campaign.

Since August 17, 2026, a budget-limited campaign stays at its target ROAS and no longer overshoots it. A target set well below the real return drags results down, and an unreachable target throttles the campaign until it stops. For the same reason, the "Limited by budget" status on a campaign with a target is no longer noise. Compare the actual return with the target first, and if the return is clearly higher, raise the target before you think about budget. Change the budget or the target, then let the campaign settle before you touch the other one.

Step 9: No assets, AI enhancements off

A feed-only campaign has no headlines, descriptions, images, or videos. Assets shift budget to Display and YouTube, and engaged-view conversions (someone watched at least 10 seconds of a video, didn't click, and bought later) make the report look better than the store's bank account does. Skip sitelinks and callouts as well, since Shopping ads don't show them.

• Turn off every AI enhancement and asset optimization. Empty asset fields aren't enough: unless you uncheck the optimization, Google generates its own text and images anyway.
• Six of these settings get turned off when you create the campaign: text, final URL expansion, image extraction from the website, image enhancement, design versions of images, and video enhancement. Newer versions of Google Ads added a seventh, videos built from web pages you point it to, so when you check, list every setting that's on, including any beyond those six. When you check them, count the types on the list. A type that nobody ever set doesn't show up when you read the settings through the Google Ads API, yet it's turned on. Five out of six looks complete, and the one most often missing is design versions of images.
• Leave the call to action set to Automated.
• Audience signals, search themes, and customer lists have no effect on targeting in this kind of campaign. Which searches the campaign serves on is driven by the product titles and descriptions in the feed.
• Turning enhancements off doesn't remove assets Google already created. In 2026 Google also generated test AI videos in campaigns without assets and with asset optimization off, and there's no campaign setting to stop that. So after launch, keep an eye on cost by channel.
• Dynamic remarketing is built into Performance Max, and remarketing taking up to about 15% of spend is normal.
• Make sure auto-apply recommendations is turned off in the account. When it's on, campaign settings change without anyone deciding. Check it monthly, because Google sometimes turns it back on.

Step 10: Products nobody clicks

A campaign can't sell a product shoppers don't click, and the title and image decide the click. The category's products with impressions and zero clicks are your fix list, before the new campaign starts spending on them. Before you blame the title, check how many impressions the product got. With Shopping CTR around 1%, no click after a few dozen impressions tells you nothing yet; after a few hundred, it's a signal.

• The title decides which searches a product shows up for. Put the main phrase as close to the start as you can, because the first 70 characters count most. Lead with the product type and its key attributes, and put the brand first only if people search for it.
• The worst title problems are product codes instead of product names, and one-word made-up names. Next come titles that leave out the product category, and typos.
• Phrases that convert but are missing from the title go to the front of it. You find them in the campaign's search terms and confirm them in Keyword Planner.
• Keep prices, dates, the store name, and promotional copy out of titles, and avoid keyword stuffing, which violates Google's policies.
• The main image works best at 1500 × 1500 pixels, with no watermark.
• You change titles through a supplemental feed, which overrides attributes of existing products, matched by item ID. Leave the primary feed as it is, without deleting or overwriting it.

The campaign doesn't have to wait for the new titles. The improved titles take effect in the campaign automatically once Merchant Center processes the supplemental feed.

Step 11: Check what the product tree actually includes

The product tree, which the Google Ads UI calls listing groups, decides which products the campaign advertises: the chosen category is included and everything else is excluded. A broken tree gives no warning. A campaign whose tree matches no products spends nothing and looks perfectly normal.

• Take item IDs from the feed, always. Google accepts IDs that don't exist without any error, and the product simply doesn't serve.
• Dead IDs in the tree push budget into Search, most of all in a campaign with assets. In one account, 438 of the 624 item IDs in the tree of such a campaign had no product performance data over 90 days, and the campaign spent 91% of its budget on Search.
• For large splits, custom labels in the feed work better than lists of item IDs, since an asset group can hold only about 1,000 listing group nodes.
• Google doesn't record product tree changes in the account's change history, so a missing entry doesn't prove that nobody touched the tree.

Step 12: Preview before approval, campaign created paused

Before anything reaches the account, it pays to see every setting in one place. The preview should be built from the exact changes that will be sent: the product tree, the budget, the status, the target ROAS or its absence, presence-only targeting, AI enhancements off, and a label with the date of the change.

• Create the campaign paused. Turning it on is a separate decision you make after reading the preview.
• The brand list, the negative keyword lists, and the category exclusion in the existing campaigns are separate changes, each with its own preview. Each one touches a different campaign or a different scope, so each gets approved on its own.
• Any change to the parameters after approval means a new preview and a new approval.
• When you look back a month later, a dated label lets you separate the effect of this change from everything else that happened in the account.
• After you create the campaign, check it again: conversion goals, locations and location exclusions, attached lists, and the tree. The preview shows what was supposed to happen, and the check shows what actually did.

Step 13: The first weeks: what to check and what to leave alone

For the first three days, check only that the campaign spends and serves. Ad approval in a new campaign takes up to 48 hours, and the campaign needs two or three days to ramp up. If there are no impressions after a day, keep the old campaigns on and look for the cause in product disapprovals, the Merchant Center link, and the tree.

• Measure a new campaign's progress by the number of conversions it has collected. Maximize conversion value needs about 15 conversions to work, and a target ROAS gets added after 50 to 60. Wait until the campaign has logged 50 conversions since the last change before you touch the target or budget again.
• Evaluate results over a window that starts at launch, minus roughly the first 14 days of learning. If fewer than 10 days are left after that, the honest answer is "too early to tell."
• Judge the category as a whole as well: sales of the same products before and after the move, over windows of equal length, with room for conversions that arrive late. The new campaign's ROAS alone won't tell you whether the account sells more or whether sales just moved from one campaign to another.
• At a 1% Conv.Rate, expect the first sale at around 100 clicks, and don't worry until you're past 200.
• ROAS for the most recent days reads low because conversions are reported with a delay. The lag affects conversions and ROAS, while impressions and clicks are accurate right away.
• Check where the campaign's results really come from by looking at the share of brand searches in its search terms and at its cost by channel.
• During this period you steer with negatives and the feed. Review search terms starting the day after your last exclusion, or you'll keep excluding the same things.
• If the campaign still has no conversions after the learning period of two to four weeks and, at a Conv.Rate around 1%, has collected more than 200 clicks, move the products back to the campaign where they sold before.

When the new campaign hits its target without clearly beating it and uses its full budget, raise the budget by 10 to 20%, once at least 50 conversions and at least a week have passed since the last change. If the return is clearly above the target, raise the target first. Performance Max doesn't report impression share lost to budget, so budget utilization is what tells you there's room.

Rules at a glance

• Before you build the campaign, count the category's conversions. With a handful a month, a separate campaign has nothing to learn from.
• Keep products with different margins out of one campaign. Break-even ROAS = 1 ÷ margin.
• Take the scope from Merchant Center, and read "not eligible" together with its reason.
• Check where the products sell today. Products shouldn't overlap between two Performance Max campaigns. Include a moved product in the new campaign first, and exclude it from the old ones only after that.
• Exclude the brand from non-brand campaigns with a brand list plus phrase match negative keywords, and keep the exclusion off the account level. Brand typos stay in the brand campaign.
• A new campaign inherits no negative keyword lists and no brand list. Attach them right away and check that they aren't empty.
• Location on presence only, checked against the source campaign for both targeted and excluded locations.
• Check the conversion goals after creation. A new campaign starts with the account-default goals, and for a store the primary goal is the purchase.
• Base the budget on the products' actual spend, and say whether it's a shift or an increase. Set a target ROAS at launch only when the category has sales history and brings in at least 50 conversions a month. One variable at a time.
• No assets, with AI enhancements, asset optimization, and auto-apply recommendations off. After launch, watch cost by channel.
• Fix products without clicks by improving titles and images through a supplemental feed. The campaign doesn't need to wait for those fixes.
• Take item IDs only from the feed, and check the tree in the preview.
• Create the campaign paused, and add lists and exclusions as separate changes with their own previews. After creation, check that the account matches what you approved.
• Judge a new campaign by its conversion count, over a window that leaves out the first 14 days, and judge the category as a whole, before and after the move.