Google Ads

The Google Ads Search Term Mistake That Makes Your Account Perform Worse After a Cleanup

Joey Sew · April 2026 · 8 min read

You open your search terms report. You sort by spend. You find a keyword that burnt $340 and converted zero times.

You add it as a negative. You feel productive.

You just made your account worse.

This happens in almost every Google Ads account we audit — not because the advertiser is careless, but because the analytical method they are using (term-by-term analysis) is the wrong tool for the job. It produces confident-looking answers to the wrong question.

Here is what you should be doing instead.

The problem with evaluating search terms individually

A search term does not exist in isolation. It belongs to an intent group — a cluster of searches that all represent the same underlying buyer behaviour, regardless of the specific words used.

When you evaluate a term on its own, you are asking: did this specific string of words convert?

The right question is: do people who search this way buy products like mine?

These are different questions. They produce different answers. And they lead to completely different actions.

What is an intent cluster?

An intent cluster is a group of search terms that share the same buyer intent — the same reason the person opened Google.

Consider a DTC brand selling premium running shoes. Their search term report might contain:

• "best running shoes for beginners" — purchase intent, comparing options before buying
• "running shoes under $200" — purchase intent, price-conscious buyer
• "running shoes for flat feet" — purchase intent, specific need
• "how to tie running shoes so they don't come loose" — informational, not a buyer
• "running shoe size guide" — research intent, not ready to buy
• "Fleet Feet near me" — stockist intent, wants a physical store

The first three belong to the same cluster: high-purchase-intent product research. A shopper who searched any of these three terms is in buying mode. If your cluster-level ROAS across all three is 4.2x, that is the signal that matters — not whether the third term had two conversions versus the first term's eight.

The last three belong to completely different clusters. They will structurally never convert on an ecommerce site — not because your ads are weak, but because the intent behind those searches is incompatible with clicking Add to Cart.

The $340 negative keyword mistake — illustrated

Back to the term that burnt $340 with zero conversions.

Imagine it is "best running shoes for half marathon training."

Evaluated individually: $340 spend, 0 conversions — obvious negative keyword candidate.

Now group it with its cluster — all high-purchase-intent running shoe searches:

• Total cluster spend: $2,840
• Total conversions: 47
• Cluster ROAS: 3.8x

The cluster is profitable. The term had a bad month — low statistical volume, no conversions by chance. Every campaign with low-traffic terms will have months like this.

If you block "best running shoes for half marathon training," you fragment your best-performing intent cluster. You reduce the data available to Google's algorithm. You potentially decrease budget flowing to terms that were actually converting.

You took a confident action that made your account measurably worse.

The terms that deserve blocking — and why

The correct reason to block a search term is not zero conversions. The correct reason is structurally incompatible intent.

Stockist intent: "running shoes near me," "running store Sydney," "Fleet Feet locations." These searchers want to try shoes on in person. They will never convert on an ecommerce store regardless of landing page, offer, or ad copy. Block the entire intent cluster.

Informational intent: "how to break in new running shoes," "running shoe lacing techniques," "what drop should my running shoes have." These people are researching a topic, not shopping for a product. Block the entire cluster.

Wrong-product intent: "running shoe insoles," "running socks," "compression sleeves for running." Adjacent categories you do not sell. Block them.

Competitor brand intent: Searches for specific competitor brands. Unless you are deliberately running conquest campaigns, these rarely convert at acceptable ROAS.

In every case, the block decision is driven by understanding why the person searched that way — not by reading last month's conversion column.

How to build intent clusters from your own data

Step 1: Export your search term report for the last 90 days.

90 days gives enough data for low-frequency terms to appear at all. 28 days misses them.

Step 2: Sort by spend descending.

You want the highest-cost terms getting attention first. Conversion rate is not the sorting criterion.

Step 3: Group every term into one of five intent types:

• Brand intent — searches containing your brand name
• Product-specific intent — searches for your product category with buying signals
• Generic category intent — broad category searches without clear purchase intent
• Informational intent — how-to, guides, tips, comparisons
• Stockist / wrong-product / competitor intent — structurally incompatible

Step 4: Calculate ROAS at the cluster level.

Sum conversion value and sum spend across all terms in each cluster. Cluster ROAS = cluster conversion value ÷ cluster spend.

Step 5: Make decisions at the cluster level.

If the cluster ROAS is strong, protect every term in it — even the zero-conversion ones. If the cluster intent is structurally wrong, block it entirely regardless of conversion history. If the cluster ROAS is weak and the intent is plausible, investigate the landing page — not the terms.

What changed with PMax search terms in 2025

Performance Max campaigns historically hid the vast majority of search term data — advertisers could only see a limited sample of queries. Google changed this in March 2025, making PMax search terms visible in the standard search terms report alongside Search campaign data.

This is a meaningful improvement in transparency. However, PMax search term data remains more limited than Search campaign data in granularity and actionability. The cluster analysis framework applies equally to PMax terms when visible — the same five intent categories, the same cluster-level ROAS calculation.

One important practical update: Google also increased PMax negative keyword limits from 100 to 10,000 per campaign in late 2025 and added native negative keyword management directly in the Google Ads interface — no support requests required. This means the cluster-level blocking decisions you make for PMax are now far easier to implement than they were previously.

When visible PMax search terms are predominantly informational or stockist intent, this is a structural signal about your audience signals and asset group targeting — not just a negative keyword problem. PMax search term issues are most effectively addressed at the campaign architecture level first, then reinforced with negative keywords.

Why this produces better outcomes

Term-by-term analysis optimises for a clean-looking report. Cluster analysis optimises for account performance.

The clean-looking report feels productive. It shows you taking action. It generates a satisfying list of negative keywords.

Cluster analysis is slower. It requires categorisation work before any decisions can be made. But the decisions it produces are grounded in buyer psychology rather than statistical noise — and they hold up over time rather than needing re-evaluation every month when the conversion columns shuffle.

The best-performing Google Ads accounts do not have the most negative keywords. They have the most accurate model of why their customers search.