How to Analyze a Search Terms Report in 20 Minutes

Why sorting by conversions hides the expensive problems

Sorting the search terms report by conversions puts the terms that worked at the top. Everything that didn't convert lands at the bottom in one long tie. In that tie, a $600 term looks the same as a $0.40 term. Expensive terms whose conversions are still on their way look exactly like terms that will never convert.

Most guides on this topic give some version of the same advice: filter for zero conversions, look for irrelevant terms, add negatives. That advice isn't wrong. It just has two gaps.

First gap: conversions are reported against the click date. Google Ads credits a conversion to the day of the ad click, not the day the sale happened. Say your buyers usually take two or three weeks to convert. Then clicks from the last 14 days will always look worse than they really are. A report sorted by conversions shows those recent clicks as failures. Any rule like "zero conversions after 20 clicks, block it" will block terms that are still working.

Second gap: zero-conversion filters ignore cost. Take an account with 4,000 search terms in a month. Thousands of them might have zero conversions, and most of those cost a few dollars each. Read them row by row and you spend an hour on terms worth less than one bad term near the top. In the accounts we audit, a few dozen terms usually carry a large share of search-term spend, often somewhere between a third and two thirds. The exact share depends on how broad the targeting is. Either way, the money is at the top of a cost sort, and the conversion sort doesn't show you the top.

So the order below starts with cost. Then it reads intent. Then it deals with timing separately. Here's how 20 minutes splits up:

Before you sort: pick a date range longer than your conversion lag

Choose a date range at least twice as long as your typical time from click to conversion. For most lead-gen and e-commerce accounts, 30 to 90 days works. A last-7-days view is fine for spotting new junk. It isn't enough to judge whether a term converts.

If you don't know your lag, check before you analyse anything. In Google Ads you can segment conversions by days to conversion. That shows how many conversions arrive on the same day and how many arrive one, two or four weeks later. A plumber taking emergency calls might see almost everything on day zero. A B2B software company with a demo-then-trial funnel might see most conversions arrive two to five weeks after the click. Those two accounts need very different date ranges.

Then set your columns. You need:

• Search term
• Match type (how the search term matched your keyword)
• Added/Excluded (so you don't redo work that's already done)
• Campaign, Ad group and Keyword (which keyword brought the term in, so you know where the fix belongs)
• Clicks, Cost, Conversions, Conv. value

Remove anything else. You aren't building a dashboard. You're making a few dozen decisions.

One more check: the report doesn't show every term you paid for. Google hides terms that too few people searched, and the hidden share can be large in niche accounts. If the costs you can see add up to much less than campaign spend, read Google Ads search terms hidden by privacy, and what to do before you draw conclusions from what's visible.

Step 1: Sort by cost and bucket the top of the list

Sort by cost, highest first. Read down until the terms cost less than about half your target CPA, or until you've read 50 rows, whichever comes first. Each term goes into one of four buckets. This pass takes about seven minutes, because you're reading words, not working out ratios.

Here's an example: an accounting software company with a target CPA of $150, looking at the last 60 days.

Keep

These terms converted at or near your target, like "best invoicing app small business" at $140 per conversion ($280 ÷ 2). Move on. If one is a strong term that came in through a broad keyword, you might add it as its own exact keyword. That's a keyword decision, not a negative one.

Keep, but fix something else

"accounting software for contractors" cost $1,200 for 6 conversions, so $200 each. That's above the $150 target, but the term is clearly the right buyer. Blocking it would throw away six customers. The fix is somewhere else: the bid, the ad, the landing page or the offer. If a term converts but the leads are poor, that's often a page problem, not a query problem. Google Ads low quality traffic: bad clicks or bad pages? shows how to tell the two apart. And if a few terms like this are pulling your whole CPA up, CPA too high in Google Ads? Check search terms first covers that pattern.

Block

"free accounting software", "quickbooks login", "bookkeeper jobs remote" and "how to do payroll" will never become customers, however long you wait. One is looking for free software. One is an existing customer of a competitor trying to sign in. One is a job seeker. One wants a tutorial. Their intent is in the words, so no amount of extra data changes the answer. Together they cost $970 in 60 days ($430 + $310 + $140 + $90).

Be careful with "free". Some people who search "free trial accounting software" are real buyers. Blocking "free" and "cheap" searches without losing buyers explains how to block the freeloaders without blocking the trials.

Watch

"accounting software comparison" cost $520 with no conversions. That's more than three times the target CPA, so the conversion-sort habit says block it. But the words describe a buyer comparing options, and comparison shoppers are usually the slowest to convert. Don't block it yet. It goes on the watch list in Step 3.

What decides the bucket is cost plus the intent in the words. The conversion count alone doesn't decide it. A term with zero conversions and buyer words goes to Watch. A term with zero conversions and non-buyer words goes to Block, even if it only cost $15.

Step 2: Read intent in the words, not the rows

After the top of the list, stop reading row by row. Filter the search term column for words that show the wrong intent, one word at a time. One filter like "contains jobs" can catch 40 cheap terms in a single pass. Reading those 40 rows one at a time would take ten minutes.

Use the "Search term contains" filter for each word below and scan what it returns. These are the word groups that most often show a searcher who will never buy:

• Jobs and careers: jobs, salary, hiring, career, internship, resume
• Learning, not buying: how to, course, training, certification, tutorial, pdf, definition, what is
• Existing customers: login, sign in, customer service, phone number, cancel, refund
• Price signals that rule out your offer: free, cheap, used, second hand, rental (only if you don't offer these)
• Wrong product or place: a competitor's product name plus "login" or "support", a city you don't serve, a product category you don't sell

Two rules keep this from going wrong.

Read the results before you block the word. "Course" is junk for a plumber. It's the main keyword for a golf course booking site. "Cancel" is junk for most advertisers, but a lawyer who handles contract cancellations wants it. Every word on that list has an account where it means a buyer. If you'd like starting lists sorted by trade, negative keyword lists by industry: where to start has them, with the same warning.

Block the pattern, not the 40 rows. If "jobs" brings back 40 different job-seeker terms, you don't need 40 exact negatives. One phrase negative, "jobs", covers them all and covers next month's versions too. That's where the time savings come from. To decide between phrase and exact, and to avoid blocking more than you meant to, see add negative keywords from search terms without guessing.

While you're filtering, check the Match type column too. If many bad terms show up as close variants of exact keywords, the problem isn't one query. Google's idea of "exact" is wider than yours. Close variants in Google Ads and why exact isn't exact explains what that column is telling you.

Step 3: Build a watch list for expensive terms that haven't had time

This is the step most search term reviews skip. Some terms cost more than your target CPA, have zero conversions, and still look like buyers. Don't block them. Write them down with the date and their cost, and check again once the conversion lag has passed.

To find them quickly, add two filters: cost greater than your target CPA, and conversions less than 1. In the example account, that's every zero-conversion term above $150. Take out the ones you already blocked in Steps 1 and 2. What's left is the watch list.

Then judge each one by when its spend happened. Narrow the date range to the most recent period equal to your conversion lag, say the last 21 days for an account with a three-week lag:

• Most of the cost falls in that recent window: too early to judge. Its conversions may still be coming. Check again in three weeks.
• Most of the cost is older than the lag and there are still no conversions: the conversions had time to arrive and didn't. Now you can block it, or fix the ad group that's showing for it, with a clear conscience.

Back to "accounting software comparison". Say $380 of its $520 was spent in the last 21 days. That's too recent to judge. Say instead that $480 was spent five to eight weeks ago and nothing converted. Then it's a block, or a sign you need a comparison page instead of sending those clicks to the homepage.

A note for slow-cycle accounts: if your sales cycle runs past a month, the "Conversions" column for any short date range will mislead you. The "Conversions (by conv. time)" column credits conversions to the day they happened rather than the day of the click. Comparing the two columns helps you see how much is still in the pipeline. It doesn't tell you which term produced it, so the date logic above still matters.

The watch list doesn't need software. A note with five lines works:

accounting software comparison: $520, 0 conv, mostly last 3 weeks. Recheck 14 March.

Step 4: Add the negatives at the right level

Blocking a term is quick. Blocking it in the right place is where reviews go wrong. Use the Campaign and Ad group columns you added earlier. Then match the scope of the negative to how widely the term is wrong.

• Wrong for the whole business ("jobs", "login" for a competitor, "how to do payroll"): put it in a shared negative keyword list applied to every Search and Shopping campaign. Standard Shopping accepts shared lists, just like Search.
• Wrong for one campaign only: say "payroll software" is junk in your invoicing campaign but is the main product of your payroll campaign. Add it at campaign level there, not in a shared list. Account-level negative keyword lists vs campaign-level covers when each scope is right.
• Right term, wrong ad group: add an ad group negative so the term goes to the ad group that has the right ad and landing page.

Performance Max needs separate handling. It accepts up to 10,000 campaign-level negative keywords but doesn't accept shared negative keyword lists. Your shared "jobs" list protects your Search campaigns and does nothing for PMax. You have to add those terms to each PMax campaign directly. Search terms for Search, Shopping and Performance Max all appear in Google's campaign_search_term_view. But the PMax view in the interface is more limited than the Search one. What the PMax search terms report actually shows explains the gaps.

Two cautions before you save:

1. Be careful with broad match negatives. Broad, phrase and exact are the three negative match types, and broad negatives behave unpredictably across match types. For most patterns from Step 2, a phrase negative is the safer choice.
2. Check that a new negative doesn't block one of your own keywords. A phrase negative "software comparison" will quietly stop a keyword you're bidding on. Google won't always warn you. Negative keyword conflicts Google Ads won't flag shows how to check.

Then write one line in a change log: date, what you blocked, at which level. Next time a client asks why traffic dropped, or a good term stops showing, you'll know where to look. If you got one wrong, how to undo negative keywords in Google Ads covers the reversal.

What this triage misses

The triage above covers what the report shows. It doesn't cover what the report leaves out, and you should know where those gaps are.

Hidden terms. The privacy threshold means part of your spend isn't attached to any visible search term. You can't add negatives for a term you can't see. You can only spot patterns in the terms you can see.

Terms with impressions but no clicks. A cost sort pushes these to the bottom because they cost nothing. Usually that's fine. When it isn't, see Google Ads search terms with no clicks, and when to block.

Display, Video and App. Those campaign types report no search terms. If a large part of your spend runs through them, this review doesn't touch it.

Conversion tracking you can't trust. Every bucket above assumes the Conversions column is accurate. If it counts page views or duplicate form fills, a term can look like a winner when it isn't. Fix the tracking first. Otherwise the triage will rank your terms by a wrong number, and do it confidently.

Twenty minutes is for a routine review. The first review of a neglected account takes longer, because there's a backlog and the patterns in Step 2 haven't been blocked yet. After two or three passes, most of the wrong-intent words are already blocked and the weekly review gets quicker.

Where PPCleaner fits

PPCleaner does Steps 1, 2 and 4 every night across Search, Shopping and Performance Max. It reads the search terms you paid for, judges which ones were never going to convert, and adds them as negatives. That includes the campaign-level PMax negatives that shared lists can't reach. It never blocks a term that converted in the last 90 days. It doesn't touch budgets, bids, ads or keywords. Every block can be undone in one click for 90 days.

What it doesn't do is the "keep, but fix something else" bucket. A term with a $200 CPA against a $150 target is a buyer you're paying too much for, and no negative keyword fixes that. That decision is about your bid, your landing page or your offer, and it stays yours. It also can't block terms Google hides from the report. The point is to spend your 20 minutes on decisions that need judgement, instead of filtering for "jobs" every week.