Reading a Google Ads Search Terms Report Without Missing Waste

The one tactic that catches most wasted spend: sort by cost, then read the report in two passes instead of one.

August 23, 2026 · QueryCut

To catch wasted spend fast, sort the search terms report by cost descending, then read it twice: once for terms that are obviously off-intent regardless of spend, and once for terms that converted poorly relative to what you paid. Sorting by cost alone misses a huge category of waste (dozens of cheap junk terms that each cost a dollar), so the two-pass structure is what actually closes the gap.

The default view hides more than it shows

Open Campaigns > Insights and reports > Search terms and the table loads in whatever order Google feels like. Most people sort by cost and start cutting from the top. That catches your single most expensive mistake, but it systematically misses the long tail: forty terms that each spent 90 cents on a query you would never bid on. Sorted by cost, those sit on page three and you never scroll there.

The mechanism to understand is that a search term is not a keyword. Your keyword is what you bid on. The search term is the actual query a person typed, matched to your keyword by Google's matching layer. With broad match and even phrase match, that layer expands aggressively: it uses your landing page, other keywords in the account, and past query behavior to decide what counts as 'related.' The report is your only window into what the matching layer actually did with your money. Waste is not one big leak; it is the sum of the matcher's small guesses that went sideways.

So the first move is not to sort. It is to widen the date range to at least the last 30 days (short windows starve you of data on cheap terms) and add columns for Conversions, Cost / conv., and Conv. rate. Then apply a filter: Cost > 0. Now you are looking only at terms that actually took money, which is the only population where waste can exist.

Pass one: intent, ignoring the numbers

The first read is a pure relevance scan. You are not looking at cost or conversions yet. You are asking one question per term: would I ever knowingly pay for this query? Scan for the classic waste signatures. Job and career queries ('[your product] jobs', 'careers'). Free and DIY intent ('free', 'how to', 'template', 'diy'). Competitor names if you are not intentionally bidding on them. Wrong-product modifiers (someone selling commercial HVAC catching 'window unit'). Location mismatches. Informational suffixes like 'reviews', 'wiki', 'reddit', 'meaning', 'definition'.

Do this pass regardless of spend, because a term that spent 40 cents today is a match-type pattern that will spend 40 dollars over a quarter. When you find an off-intent term, look at the word driving it and add that word as a negative, not just the exact phrase. 'free' as a phrase-match or broad negative kills 'free trial', 'free download', and 'is it free' in one move. Cutting only the exact term you saw leaves the pattern alive.

One caution: 'free' is exactly the kind of negative that backfires if you sell a freemium product or run a free-trial offer. Blanket-negativing it would block your best converters. This is why pass one is a human read and not a rule. You are judging intent against your specific offer, which is the one thing Google's matcher cannot do reliably.

Pass two: spend versus return, with real thresholds

The second read is where numbers matter. Now sort by cost descending and go top-down looking for terms that are on-intent but underperforming. The trap here is cutting too early. A term with zero conversions is not proof of waste; it might just be under-sampled. A single term that has taken three clicks and no conversion tells you almost nothing, because one missed conversion at that volume swings the whole read.

Use this heuristic: wait for roughly 100 clicks on a term before you trust a zero-conversion verdict. Below that, a converter can look like a loser purely by chance, and the cost of waiting a few weeks is usually smaller than the cost of cutting something that was about to pay off. Above that, if the term has spent more than 3x your target cost-per-acquisition with zero conversions, cut it. Between 1x and 3x CPA with zero conversions, watch it; between those bounds you are still inside normal variance for a term that converts occasionally.

Recency is the second dimension people forget. A term that spent heavily but whose last click was 45 days ago may already be starved by your negatives or by bid changes. Cutting it does nothing but tidy the list. Prioritize terms that are both expensive and still active in the last 7 to 14 days, because those are the ones actively bleeding right now. Add a filter on Clicks in the recent window to surface them.

Where the standard advice backfires

The obvious move (add every non-converting search term as an exact-match negative) is the most common way to quietly wreck an account. Three failure modes. First, over-negativing at the phrase or broad level can block queries you want: a plumber who negatives 'repair' as a broad negative to kill 'appliance repair' also blocks 'pipe repair'. Always check what else a proposed negative would catch before you commit it.

Second, negatives collide with your keywords. If you negative 'cheap' but also bid on 'cheap flights', the negative wins and your keyword goes dark with no warning in the interface. When adding negatives, search your own active keyword list for the word first. If it appears, use a longer, more specific negative phrase instead.

Third, aggressive negativing on a broad-match campaign starves Smart Bidding of the exploration it needs. Broad match plus a value-based bid strategy is supposed to find new converting queries. If you negative every term that has not converted in its first handful of clicks, you cut off the experiments before they resolve. On those campaigns, raise your click threshold and your patience: judge terms on cost against target, not on the mere absence of a conversion in a small sample.

Making it a repeatable pass, not a heroic cleanup

The reason waste accumulates is that nobody reads the report until spend hurts, and by then there are thousands of terms. The fix is cadence over intensity. On an account spending steadily, a weekly 15-minute pass beats a quarterly deep clean, because you are reviewing a small, recent slice each time instead of drowning in history.

Structure each pass the same way: filter Cost > 0 over the last 7 to 14 days, run the intent scan first, then the spend-versus-return scan with your CPA thresholds, then check every proposed negative against your keyword list before applying. Keep a running note of negatives you added and why, so you can reverse a bad call when a campaign's traffic drops.

This is also the part that scales badly by hand once you are past a few hundred terms per week, which is where tooling earns its place. QueryCut reads the full term list and flags the off-intent candidates so your human judgment goes to the ambiguous calls (the 'free' and 'repair' cases) instead of the obvious ones. But the method above stands on its own: the tool changes how long the pass takes, not what a correct pass looks like.

Questions people also ask