Performance Max for Ecommerce: Exclude Low-Return Products

To exclude products from Performance Max (PMax), give them a Merchant Center custom label, subdivide the asset group’s listing group on that label, then click the green circle next to the label value you want out and select Exclude (Google Ads Help). Exclude Dogs first: products with many ad clicks and a return below their own break-even.

Last reviewed 7 October 2026; the sources and method are at the end.

The audience is PPC specialists and ecommerce owners running Performance Max on anything from a few hundred to tens of thousands of products. PMax is Google’s most automated campaign type, and its budget follows the products that already get clicks, which aren’t always the ones that make money.

We build Product Metrics, product-level optimisation software for Google Shopping and Performance Max, so the examples use its segments. The maths behind the targets is in break-even ROAS per product.

Seven ways to exclude, and only some touch products

“Exclude” means different things in Google Ads. The first four methods act on products. Brand exclusions and negative keywords act on search queries, and placement exclusions act on where ads show, so none of those three takes a product out of anything.

This guide covers the first two rows, which is what “exclude products from Performance Max” asks for. Two topics are left out because no Google help page we checked confirms them: how PMax and Standard Shopping rank against each other when they share products, and “search themes” as an exclusion tool.

How do you exclude products from Performance Max?

Subdivide the listing group on a custom label, then exclude the label value. Google’s steps are four clicks once the label exists.

The result is a small tree per asset group. Google Ads Editor’s help says a new subdivision starts with one item, Everything else (Google Ads Help).

With Product Metrics the label is already there. It writes each product’s segment to custom_label_0, so excluding Dogs means subdividing on custom_label_0 and excluding the dogs value. The label updates daily, so products move in and out of that value as their segment changes, and the exclusion follows the label once Merchant Center has read the updated feed.

Exclude Dogs first, not Drainers. Dogs have had real spend at a low return, while many Drainers have spent too little to judge. If you would rather keep Dogs running, give them their own campaign and a target set from their break-even instead.

An exclusion only reaches as far as its method

These are the limits Google states.

If you exclude on a label, re-including a product is a label change, not a campaign edit. With daily segments that happens by itself when a Dog earns its way out.

Why spend follows clicks

Performance Max serves your ads across Google’s inventory, including YouTube, Display, Search, Discover, Gmail and Maps, from a single campaign (Google Ads Help, About Performance Max campaigns). For a shop, your Merchant Center feed is one of its inputs: you hand Google the products and a goal, and it decides how to spread the budget.

In a typical set-up, every product in an asset group shares one budget and one target, so products with many ad clicks take a large share of the spend. If those clicks don’t turn into return, the budget goes to them anyway.

The chart shows the pattern in a segmented product range. Stars (high clicks, high return) and Dogs (high clicks, low return) both take around a third of the spend, yet Dogs are twice as many products. Drainers are by far the largest group and take only a small part of the budget.

Read the spend bars before the product bars. Dogs take the most spend at the lowest return, so they are the first place to look. Whether that return is a loss depends on margin, as the next sections show. Drainers look worse by count, but they cost little.

Which products is Performance Max spending on?

Open the campaign’s asset groups and go to the Listing groups tab, which reports product-level data (Google Ads Help). Sort by cost and the biggest spenders come first. Google notes that this tab can show different metrics from the Campaign tab, because it only reports product-level data.

To see whether that spend paid off, compare each product’s return with its own break-even and group the products by segment, category or label. The Performance Max page shows how Product Metrics lays this out per product. To check whether Performance Max brings sales you would not have had anyway, see incrementality testing for Google Shopping.

June 2026: product reporting got wider

Google widened product reporting in June 2026. This is its wording (Google Ads Help):

Starting June 2026, product reporting has expanded. Previously, product performance metrics such as cost and conversions were only available for products included in Performance Max ads serving on Search networks and Standard Shopping campaigns. […] This includes data for all networks in Performance Max campaigns […]. This may cause a one-time increase in metrics such as impressions, clicks, and more for Performance Max campaigns only.

Google’s page gives no guidance on comparing data across the change. Our reading is that a product’s cost and conversions before June 2026 covered fewer networks than they do now, so a before and after exclusion test that spans the change mixes two things: the exclusion and the wider reporting.

Two habits help. Keep both periods of a test after the change, and write the date of any change to a target, budget or exclusion in a log. A jump or drop in product-level numbers around June 2026 then has an explanation that isn’t your exclusion.

One account target, many margins

Products with different margins need different returns, so a Dog at 1.7 and a Star at 5.2 are judged by one number. POAS, which is ROAS × margin, puts them on one scale: 1.0 is break-even for every product, whatever its margin (POAS vs ROAS).

Whether 1.7 is a loss depends on that product’s margin. A product with a 60% margin breaks even at about 1.7. One with a 25% margin needs 4, so a 1.7 Dog sits at break-even on one product and makes a clear loss on another.

Compare each group’s return with its own break-even rather than reading the segment alone, using the calculation in break-even ROAS per product or the break-even ROAS calculator. Then set each group’s target from that number instead of from the account average.

Missing conversions make products look worse than they are, so before you judge a group, check what Full Signal Tracking sends back to Google Ads. Open the ad blocker tracking test in your own browser to see how much a blocker hides, and use the SDK debugger to decode a single tracking request.

Six segments by clicks and return

Every product can be placed on two axes: volume, measured in ad clicks, and return, measured in ROAS or POAS against your target. Product Segmentation does this with Product Metrics ML and puts each product in one of six segments. It works on ROAS right away, without margin data; connect margins to switch to POAS.

A sixth segment, Ghosts, holds products with no impressions and no clicks. Segments are stable: a product needs enough clicks before its segment changes, so products don’t jump between groups week to week. Once a Ghost’s feed is in order, Demand Gen vs Performance Max shows how to try it in a small Demand Gen test.

Exclude, split or wait: a rule per product

For each product, divide 1 by its contribution margin and compare its return over a full review window with that number. We suggest this rule:

• Exclude a product that returns less than its own break-even and has enough clicks to judge. You choose the minimum; don’t judge on a handful of clicks.
• Split a product that clears its own break-even but misses your account target. Give it a campaign with a target set from its break-even, if its group keeps enough conversions (see the data section below).
• Wait on a product with too few clicks, a recent change or a conversion delay still open, and on one bad week.

Three products show it. POAS is ROAS × margin.

Illustrative data, not a benchmark.

Products A and B return the same 1.7, yet only A loses money, because A’s margin is 25%. A single account target ROAS of 400% would treat them alike and wrongly push B out. Product C looks worse than B, but has too few clicks to say.

Before you exclude a Dog with many clicks, look at your shop’s basket data: do its buyers also buy other products? If so, excluding it may cost you those orders.

How do you structure Performance Max campaigns for ecommerce?

Three campaigns is enough to start with. Stars get a campaign of their own, Question Marks, Cash Cows and Drainers share a catch-all, and Dogs go in a campaign of their own or are left out. It makes a sensible default because each campaign stays big enough for Performance Max to learn from.

Splitting lets Stars scale without Dogs drawing on the same budget. It also lets you set a tighter target on Dogs without dragging down the products that earn.

You then filter your asset groups or campaigns on the label once, and set the targets and budgets yourself.

You also choose the period that segments are calculated over. They are recalculated daily, and 30 days is the default. A 14-day period reacts faster to a bad week, while a 30-day period evens it out.

Pick the shorter period when your range changes quickly, for example with seasonal products or promotions, and the longer one when you want fewer moves. With either, the minimum-click threshold stops a product changing segment on a handful of clicks.

That also explains how a product can stay a Star after one bad week. Over 30 days, the week is averaged in with the stronger weeks before it, and with a 14-day period it drops out sooner.

How many Performance Max campaigns should you run?

Prefer fewer, larger campaigns. smec’s State of Performance Max 2025, an analysis of more than 4,000 campaigns, says Performance Max campaigns need at least 30 conversions a month, ideally 60 or more, and that “the target ROAS accuracy improves dramatically” as monthly conversions rise. The chart on the Performance Max page groups that data by monthly conversions: the share of campaigns below their ROAS target falls as volume goes up.

Those thresholds are smec’s. The roughly 150 conversions a month is our own suggestion for a campaign you split, so that each campaign keeps well above that minimum. If splitting would leave a campaign short of conversions, keep that group in the catch-all instead.

Set a target ROAS on Maximise conversion value instead of leaving it open. Google says that without a target ROAS the strategy “tries to fully spend your average daily budget”, and that with one it behaves like a Target ROAS strategy (Google Ads Help). So with no target there is no return floor to hold low-return products back.

Choose the target for each group from its break-even (the break-even ROAS calculator gives it for any margin). Change it slowly, and judge the result after the conversion delay rather than after two days.

Reviews and the change log

Judge each group against its own break-even once a week at most, and wait for the conversion delay before you read a change. Recent days look worse than they are because some conversions arrive later, so a Tuesday review of Monday’s change tells you very little.

Keep a note of every change to targets, budgets and bid strategies, with its date. When a group’s return moves, you can then tell whether the product mix or your own change moved it. In Product Metrics, the Campaign Explorer overlays those changes on ROAS against target ROAS for Shopping and Performance Max campaigns, with a conversion delay of up to 30 days.

Product Metrics reads your Google Ads and shop data, segments your products and writes the labels you approve; which group gets which target, and which products leave the main campaign, stay your decisions. The basket-data check before excluding a Dog lives in your shop analytics.

Your first exclusion

To set a target for each group, see how to set a target ROAS from your margin; to judge the result in profit, POAS vs ROAS; and for Standard Shopping, how to optimise Google Shopping ads.

Work out break-even ROAS for your five biggest products, then check which segment they sit in: the free analysis in Product Segmentation finds your Dogs, and the Performance Max page shows the per-product view. If Dogs are a large share of your spend, that is the first exclusion to make.

How we checked this guide

Written and reviewed by Berend Vrakking, founder of Product Metrics, whose work was nominated at the Dutch Search Awards (Best Measurement Campaign 2023, Best Use of Data 2024).

Each page below was read in October 2026. The 30 and 60 conversion thresholds are smec’s; the 150 is our suggestion.

• Google Ads Help: Manage a Performance Max campaign with listing groups, About product groups and listing groups in Google Ads Editor, About Performance Max campaigns
• Google Ads Help, query and placement exclusions: Apply brand exclusions, About brand settings, Negative keywords in Performance Max, Exclude placements at the account level, Brand suitability in Performance Max
• Google Ads Help, reporting and bidding: Product reporting changes, Maximise conversion value
• Google Merchant Center Help: Manage products, Excluded destination
• Google Ads API: Listing groups in Performance Max
• smec: State of Performance Max 2025