Search terms report: how to analyse it in Google Ads

Your campaign spent £12,400 last month. The search terms report accounts for £10,850 of it. The rest went on searches Google won't show you.

That's the report in one sentence: useful, and incomplete on purpose. It lists the searches that triggered your ads, with clicks, cost and conversions for each. Google describes it as "a list of search terms that a significant number of people have used, and that resulted in your ad being shown". Terms below that bar are left out: Google says some search terms "that don't have enough query activity are omitted from the search terms report in order to keep with our standards on data privacy."

So the report can't show you every penny. What it can show you is where most of the money went, which is enough to act on. Find it under Campaigns, then Insights and reports, then Search terms.

Two columns do most of the work. Keyword shows which of your keywords matched the search, and it's hidden by default: add it from the columns menu under Attributes. Match type shows how closely the search matched that keyword. Add conversions and cost, and you have everything this page uses.

Search terms insights, a separate view that groups searches into themes, does account for the low-volume searches by grouping them into themes or "other queries", without showing the queries themselves. Use it for themes. Use the report when you need the exact words.

A search term is what someone typed. A keyword is what you bid on. Google's example: someone searches "red roses", and your ad can show because you added the keyword "roses". The search term is "red roses"; your keyword is still "roses".

The gap between the two is set by match type:

• Exact match shows your ad on searches with the same meaning or intent as the keyword.
• Phrase match shows it on searches that include the meaning of the keyword.
• Broad match shows it on searches that are related to your keyword. Google's words: these "can include searches that don't contain the direct meaning of your keywords." Broad is the default for every new keyword.

Broad match is where most surprising search terms come from. It also explains a puzzle in the report. The Match type column describes the search, not your keyword, so a broad match keyword that triggers on an identical search shows up as exact match. Read the Keyword column beside it before you change anything.

Search term analysis is the habit of reading these two columns together, then doing two things: blocking the searches that cost money and never convert, and promoting the searches that convert into keywords of their own. Doing that well moves your Cost Per Acquisition (CPA) more than most bid changes will.

Here's a month for a payroll software company bidding on "payroll software". The figures are invented to show the arithmetic. The conversion is a demo request, and the campaign as a whole spent £12,400 for 70 of them: a CPA of £177.

Step 1: check the totals

Add up the rows before you read them. Here they come to £10,850 and 62 conversions against the campaign's £12,400 and 70. The missing £1,550 and 8 conversions sit in searches the report doesn't list. Write that gap down. Any claim of the form "we wasted X%" should say whether it's a share of the rows shown or of the whole campaign.

Step 2: find the wasted spend

Filter for search terms with no conversions. Three rows, £2,170: that's 17.5% of the campaign's spend, on people looking for a job, a course or a free how-to guide. None of them were ever going to book a demo.

Then look at the expensive converters. "free payroll software" converted once for £1,520, more than eight times the campaign's CPA. If you don't offer a free plan, it belongs with the other three.

Don't be too quick with thin rows. A term with 15 clicks and no conversions hasn't proved anything yet. A workable rule: judge a term once it has spent more than your target CPA without converting.

Step 3: add negative keywords

Tick the rows in the report and choose Add as negative keyword. Two details matter here.

• The default is exact match. Google adds negatives from the search terms report as negative exact match, so excluding "payroll jobs" blocks that search and nothing else. "payroll jobs london" would still show your ad. For a word that's never relevant, add it as a broad negative instead: jobs. A negative broad match keyword blocks any search containing all of its words, in any order.
• Negatives don't match close variants. Google is explicit: negative keywords "won't match to close variants or other expansions." Excluding jobs doesn't exclude job, so add both.

For "how to do payroll uk", be narrower. A negative on how to would also block "how to choose payroll software", which might be your best search. Exclude the phrase "how to do payroll", and check every negative against the converting rows above before you save it.

Step 4: find new keywords

"payroll software for charities" converted 6 of 95 clicks at £95 each, about half the campaign's CPA. It only shows because the phrase match keyword happens to catch it. Add it as its own exact match keyword, with an ad and a landing page that mention charities, and you control its bid and its message. Six conversions is a small sample, so give it a month before you judge the new keyword.

The result: £3,690 a month freed from searches that bought one demo between them. Whether that turns into savings or into more demos depends on the next section.

Impression share answers a different question from the search terms report. The report shows where your ads did appear. Impression share shows how often they could have appeared and didn't.

Google defines three numbers for Search campaigns:

• Search impression share: "The impressions you've received on the Search Network divided by the estimated number of impressions you were eligible to receive."
• Search lost IS (budget): "The percentage of time that your ads weren't shown on the Search Network due to insufficient budget." It's available at campaign level only.
• Search lost IS (rank): "The percentage of time that your ads weren't shown on the Search Network due to poor Ad Rank in the auction."

Ad Rank is Google's score for each auction, built from your bid, the quality of your ad and landing page, the minimum thresholds, the competition, the context of the search and the expected impact of your assets. Losing on rank means losing auctions, not running out of money.

Read the three together

Google names two reasons an eligible impression is lost, budget and rank, so read the three figures as one split of the impressions you could have had. Say the payroll campaign above shows a search impression share of 58%, lost IS (budget) of 27% and lost IS (rank) of 15%. It missed about a quarter of its chances because the money ran out, and a smaller share because it lost the auction.

The trap is in the second row. When a campaign loses impression share to budget, the obvious fix is a bigger budget. But if 17.5% of the spend goes on job hunters and students, a bigger budget mostly buys more of them. Add the negatives first. In a budget-limited campaign, the £3,690 they free up goes back into the auction on searches that convert, and lost IS (budget) falls without a penny more spent.

A few things that catch people out in the data:

• Impression share is an estimate. Google says small fluctuations "don't necessarily indicate that action is needed."
• Lost IS (rank) isn't shown on the Ad groups tab if the campaign ran out of budget at any point in the date range.
• Impression share is reported separately for each campaign type and isn't added up across the account. Search partners aren't included.
• The figures update within one to two days, so yesterday's numbers may still move.

Export the search terms report as a CSV, attach it to Claude or ChatGPT, and ask which search terms are wasting money. You'll get a confident list in seconds. Whether it's right depends on things the answer won't mention unless you ask: which column it read as cost, whether it summed the rows or averaged a ratio, which threshold it used for "wasted", and whether a suggested negative would also block a search that converts.

So ask for the working. This prompt makes the model show the filters, row counts and totals behind every figure, and stop if the file doesn't contain what it needs.

I've attached a Google Ads search terms report export for [date range].
For the same period the campaign's total cost was £[X] and total conversions were [Y].
My target cost per conversion is £[T].

1. List the columns in the file and say which one you are using for search term,
   keyword, match type, clicks, cost and conversions. If any is missing, stop and tell me.
2. Total cost, clicks and conversions across all rows. Compare them with the campaign
   totals above and report the difference as spend on search terms the report doesn't list.
3. Wasted spend: list every search term with zero conversions and cost above £[T].
   State the filter, how many rows it returned and their total cost, as a share of the
   rows shown and of the campaign total.
4. Expensive converters: search terms with conversions whose cost per conversion is more
   than twice the campaign's. Show the rows and the filter.
5. Negative keyword candidates: for each, give the negative keyword, its match type
   (exact, phrase or broad), and every search term in the file that it would also block.
   Flag any that would block a search term with conversions.
6. New keyword candidates: search terms with at least 3 conversions that aren't
   already a keyword (the Keyword column differs from the search term). Show the rows.
7. Every figure you quote must come from a row or a total shown above. Don't estimate
   anything the file doesn't contain.

Step 5 matters most. A negative keyword that blocks a converting search does more damage than the waste it removes, and it's the step a model is most likely to skip.

An export goes stale. Connected through Count's MCP server, Claude can query your Google Ads data where it's kept, and each figure comes back with the query that produced it, so you can rerun next month's analysis on next month's data and check the same filters. Claude for marketing covers the wider monthly reporting setup.

Google's own advice for raising impression share: increase the budget, raise bids, improve ad quality, or narrow the regions you target. Each works, and each has a cost. Which one to reach for depends on why the impressions were lost.

Lost to budget: fix the waste before the budget. Run the search term analysis above first. Negatives on irrelevant searches free spend for the searches you want, and raise impression share where it matters without raising cost. Then, if your Cost Per Acquisition (CPA) is below target, more budget is worth buying.

Lost to rank: work on Ad Rank. Bids are one lever. Ad relevance and landing page experience are the others, and a better Click-Through Rate (CTR) feeds into how Google judges your ads. Raising bids alone tends to raise your cost per click too.

Narrowing targeting raises the percentage, not the reach. Google notes that fewer regions can raise impression share while reducing the impressions available to you. A higher share of a smaller pie isn't growth.

Don't chase 100%. The last impressions are the most expensive ones, and some of them come from searches you'd add as negatives anyway. Judge the extra spend on what it returns: Return on Ad Spend (ROAS) for an online shop, cost per qualified lead for B2B, and Conversion Rate to check the new clicks are as good as the old ones.