When One PMax Asset Group Pulls Irrelevant Traffic

Most advice about Performance Max asset groups stops at "make more of them" or "check your asset group report". Neither tells you what to do when the report shows one asset group with 60% of the spend, a $0.28 CPC and no conversions, while the other three convert fine at $1.40. That is not an asset quality problem. That is a targeting problem wearing an asset group costume, and the fix lives mostly outside the asset group itself.

Why one asset group can take most of the budget

Performance Max does not divide budget evenly between asset groups. It spends where it predicts the next conversion is cheapest, and prediction leans heavily on signals it already has — click-through rate, historical engagement, the breadth of the audience the asset group can reach. An asset group with broad imagery, generic headlines ("Shop Our Full Range", "Great Prices Every Day") and a loose search theme can be matched against far more queries than a tight one. More eligible auctions means more impressions, more impressions at low CPCs means more clicks, and the bidding system reads that volume as momentum.

Here is the trap. Cheap clicks make the asset group look efficient on the metrics the system optimizes against before conversions arrive. A hoodie retailer we would expect to see this in might have four asset groups: Men's Hoodies, Women's Hoodies, Kids, and a catch-all "Winter Essentials" group with lifestyle photos and no product specificity. Winter Essentials gets matched to "winter outfit ideas", "cold weather tips", "how to layer clothes", "hoodie drawing". All cheap. All useless. But at $0.22 a click against $1.60 for the product-specific groups, it can absorb half the daily budget before anyone notices, because the campaign-level CPA still looks acceptable — the converting asset groups are carrying it.

The campaign total hides the split. That is the whole problem. You look at a PMax campaign at $2,000 a month and a $38 CPA, decide it is fine, and never see that $900 of that $2,000 produced nothing at all. The same structural blindness we described in where Performance Max wasted spend hides applies here, one level down.

The signal that something is wrong

Three numbers, side by side, per asset group:

• Cost share — what percentage of campaign spend this asset group took.
• Conversion share — what percentage of conversions it produced.
• Average CPC.

If one asset group has a cost share more than roughly double its conversion share, and a CPC well below the campaign average, you have found it. The low CPC is the tell. An asset group that genuinely competes for buying intent pays buying-intent prices. An asset group paying a third of the campaign average is not winning better auctions — it is winning different, cheaper, less commercial auctions.

How to prove it, since Google will not show you directly

Google does not report search terms per asset group. campaign_search_term_view is exactly what the name says: search terms at the campaign level, for Search, Shopping and Performance Max. There is no asset-group dimension in it. So you cannot open a report and read "this asset group bought these queries". Anyone telling you to just check the asset group search terms is describing something that does not exist.

What you can do is build the evidence indirectly, in three ways.

Read the asset group report for cost, conversions and CPC. In the Google Ads UI, Campaigns → your PMax campaign → Asset groups, then add Cost, Conversions, Conv. value and Avg. CPC as columns. This gives you the split. It does not tell you what the bad asset group bought, but it tells you which one is bad.

Read the campaign search terms report and match by vocabulary. Open the campaign's search terms. You will usually be able to attribute queries to asset groups by language alone. "hoodie drawing" and "how to layer clothes" did not come from the Men's Hoodies asset group with product shots and a size chart. They came from the lifestyle group. This is inference, not proof, but in practice it is right often enough to act on. Be honest that it is inference — if two asset groups cover overlapping products, you genuinely cannot separate them, and you should not pretend otherwise. The limits of this report are worth knowing in detail: what the PMax search terms report actually shows covers the categorization, the thresholds and the terms you never see.

Pause and watch. The brute-force method. Pause the suspect asset group for two weeks and watch three things: campaign spend, campaign conversions, and the search terms list. If spend drops and conversions do not, you were right. If the junk queries disappear from the search terms report in the same window, you were right twice. This costs you two weeks of learning and some volume, and Google will warn you about limited asset coverage. Take the warning for what it is — a nudge toward spending more, not a diagnosis.

Why "just add more asset groups" is not the answer

Several guides recommend splitting into more asset groups as a general improvement. More asset groups give you more granular reporting, which helps. But adding a fifth asset group does not stop the second one from pulling junk. It gives the budget one more place to go, and if the junk-pulling group still has the cheapest clicks, it still wins. Granularity is a diagnostic tool, not a fix. The fix is removing or tightening the asset group, and blocking what it bought.

Negative keywords in PMax are campaign-level, and that changes the plan

You cannot attach a negative keyword to an asset group. Performance Max takes negative keywords at the campaign level only — up to 10,000 of them — and it does not accept shared negative keyword lists. Standard Shopping takes both negatives and shared lists, like Search does. Performance Max is the exception.

This has a practical consequence people skip past. If your "Winter Essentials" asset group brought in "how to layer clothes", you block the phrase at campaign level, and it is blocked for every asset group in that campaign. Usually that is fine — none of your asset groups want that query. But consider a campaign with a "Hoodie Gifts" asset group that genuinely converts on "gift ideas for teenagers" while the lifestyle group burns money on "gift ideas" generally. Block gift ideas as a phrase negative and you kill both.

So before you block, check the converting side. Three rules we apply:

1. Never block a term that converted recently. PPCleaner never blocks a term that converted in the last 90 days, regardless of how irrelevant it looks. A term that produced a sale is not irrelevant, whatever your instinct says.
2. Prefer exact negatives when asset groups overlap. [gift ideas] blocks one query. "gift ideas" blocks a family. When one asset group needs part of that family, exact is the only safe choice, even though it means more work.
3. Avoid broad negatives. Broad negatives behave unpredictably across match types and are the most common cause of a block being wider than intended. Negative keyword match types don't work like positives explains why the mental model you carry over from positive keywords misleads you here.

If the conflict is real — one asset group needs terms another asset group is wasting money on — the campaign-level negative cannot resolve it. The only structural answer is to move the asset group into its own campaign, where it gets its own negative list and its own budget. That costs you learning time and splits your conversion signal, which is a genuine price. Do it when the overlapping spend justifies it, not as a reflex.

The four asset group types that reliably pull junk

Across accounts we audit, the asset groups that drag a PMax campaign down tend to fall into a small number of shapes. Not a law, but a pattern worth checking against.

The brand or "about us" asset group

Built with company photos, the logo, headlines about heritage and service. No products, no prices, no specific intent. It matches against everything from competitor names to "[your brand] careers" to "[your brand] returns policy". Job seekers are a particularly expensive subset — see job seekers clicking Google Ads for the term list. Brand exclusions help with some of this, but not all of it; Performance Max brand exclusions and what they miss sets out the gaps.

The category catch-all

"All Products", "Our Range", "Shop Everything". Usually created so the campaign covers inventory that does not fit the other groups. Because its listing group is wide open and its copy says nothing specific, it matches almost anything in the vertical. If you run Shopping inventory inside PMax, this is also where the listing group leaks are: how listing groups hide wasted spend in Performance Max goes through that side.

The top-of-funnel asset group

Built deliberately for awareness, with soft headlines and informational search themes. The intent is reasonable — reach people earlier. The result in a conversion-optimized campaign is that it consumes budget from a pot judged by CPA, and it will always lose that comparison while winning the impression race. If you want top-of-funnel, give it its own campaign and its own budget. Inside a conversion campaign it is a leak.

The new asset group still in learning

Recently added, no history, so the system explores. Exploration means cheap, broad queries. This one is temporary and you should not panic-pause it in week one. Give it three to four weeks. If cost share is still double conversion share after that, it is not learning, it is leaking.

What to fix first, in order

Work from cheapest action to most expensive. The order matters because each step changes what the next step sees.

One. Block the obvious junk at campaign level. From the campaign search terms report, pull every term with meaningful cost and zero conversions that no asset group could want: informational queries, "how to", "diy", "free", "jobs", "salary", "meaning", "wikipedia". These are safe because no asset group in a commercial campaign benefits from them. Use phrase for families, exact for anything borderline. If "cheap" and "free" feel borderline in your vertical — and they often are — blocking free and cheap searches without losing buyers works through when those modifiers still signal a buyer.

Two. Tighten the search themes on the offending asset group. Search themes are suggestions to the system, not keywords. They widen reach; they do not restrict it. Removing a loose theme like "winter clothing" reduces how far the asset group reaches, but it does not create a boundary. The relationship between the two levers is covered in search themes vs negative keywords in Performance Max — the short version is that themes open doors and only negatives close them.

Three. Rewrite or narrow the assets. Replace lifestyle imagery with product imagery. Replace "Great Prices Every Day" with "Men's Fleece Hoodies from $45". Specific assets match specific intent. This is slower-acting than a negative keyword and you will not see the effect for two or three weeks.

Four. Pause the asset group. If cost share is still double conversion share after the first three steps, remove it. Keep the assets — you may want them in a separate campaign later.

Five. Split into a separate campaign. Only if the asset group has a real, different job and you have enough conversion volume to support two campaigns. Below roughly 30 conversions a month at campaign level, splitting usually hurts more than the waste you were trying to stop.

After each step, wait. Changing four things in one afternoon means you learn nothing about which one worked. How often should you check search terms in Google Ads sets out a cadence that does not turn into daily fiddling.

What this costs you, honestly

Three things are worth stating plainly, because most articles on this topic do not.

You will sometimes block a term another asset group wanted. Campaign-level negatives cannot distinguish. The mitigation is exact match on borderline terms and never blocking a converter — but mitigation is not elimination. If you block "winter layering" and three months later a different asset group would have sold a fleece against it, you will not see the sale that did not happen. That is the structural cost of PMax not supporting asset-group negatives. Anyone who tells you the trade-off does not exist has not run the campaign.

Pausing an asset group reduces reach, and Google will tell you so loudly. The optimization score will drop, the recommendations tab will nag, and the warning about insufficient asset coverage will appear. None of that means you were wrong. Google Ads optimization score and what it ignores covers why the number rewards breadth over profit.

Some of the waste is not in search terms at all. Performance Max spends on Display and YouTube placements, and campaign_search_term_view reports nothing for Display, Video or App. A negative keyword cannot touch a placement. If your bad asset group is leaking on content placements rather than queries, you need exclusions, not negatives — Performance Max placement exclusions, honestly explains what you can and cannot exclude. Expect to use both tools, and do not assume a negative keyword list solved a problem it was never able to reach.

Keeping it from coming back

An asset group that pulled junk once will pull it again as soon as inventory changes, a new product line is added, or someone drops in a fresh set of lifestyle images. The query mix in Performance Max is not stable — it drifts, and it drifts fastest right after any change to assets, feed or budget.

Two habits keep it under control. First, check the asset group cost-vs-conversion split monthly, not quarterly. It takes four minutes and it is the only place the imbalance is visible. Second, keep the campaign-level negative list current rather than doing one big cleanup a year. The 10,000-negative ceiling is generous; you will not hit it with normal maintenance, and worrying about how many negative keywords are too many matters less than worrying about the ones you never added.

This is the part PPCleaner handles. It reads the search terms your campaigns actually paid for — Search, Shopping and Performance Max — judges which ones were never going to convert, and adds them as campaign-level negatives every night. It does not touch budgets, bids, ads or keywords, so it will not pause your bad asset group for you. That judgment stays yours. What it removes is the slow leak between your monthly reviews: the terms the lifestyle asset group picked up on a Tuesday that would otherwise sit there until you next opened the report. And if a block turns out to have been wrong, you can undo it in one click for 90 days.