Reading a search terms report properly

The keywords tab shows what you bid on. The search terms report shows what people actually typed before your ad appeared. With broad match and automated bidding doing more of the work, the distance between the two keeps growing, and that distance is where search budgets leak.

Most accounts look at search terms only when something has gone wrong. A short weekly routine catches the leaks while they are small.

Before you start

• Pick the right date range. The last 7 days catches new problems; the last 30 days gives enough volume to judge terms that convert slowly. Alternate weekly, or use 30 days on a rolling basis for smaller accounts.
• Know your target. Write down the cost per acquisition or ROAS each campaign needs. Without it, every judgement becomes a matter of taste.
• Keep a log. A simple sheet of terms excluded, the date and why. It stops you re-adding a negative someone removed for a reason, and lets you undo mistakes.

The fifteen-minute routine

Minutes 1–5: sort by cost and look for the obvious

Sort descending by cost and read the top fifty terms. You are looking for intent that does not match what you sell:

• Job seekers: “careers”, “jobs”, “salary”, “internship”.
• Existing customers: “login”, “track order”, “customer service”, “cancel”.
• Bargain or free intent you cannot serve: “free”, “template”, “DIY”, “second hand”, if those do not apply to you.
• The wrong product: a different meaning of the same word, or a category you do not stock.
• Places you do not serve.

These become negatives straight away.

Minutes 5–10: filter for spend without results

Filter for terms with zero conversions and cost above your target cost per acquisition. Each one has already spent more than a customer is worth without producing one. Read them individually. Some are relevant terms that simply have not converted yet; leave those and check again next week. The rest are candidates for negatives.

Then look at terms with conversions but a cost per acquisition far above target. These are often too broad: “running shoes” instead of “trail running shoes women”. Rather than excluding them, consider whether a more specific keyword with its own bid would serve them better.

Minutes 10–13: find the winners

Now sort by conversions. Terms that convert well but are only matching through a broad keyword deserve their own keyword, usually as exact match. That gives you control of the bid and the ad copy for the people who are most likely to buy, and makes their performance visible on its own row.

Minutes 13–15: check brand leakage

Search for your own brand name in the report for non-brand campaigns. If brand searches are being picked up by generic campaigns, they are inflating those campaigns' results and usually paying more than the brand campaign would. Add the brand as a negative in non-brand campaigns so each campaign reports on what it is supposed to do.

Adding negatives without breaking things

• Choose the match type deliberately. A phrase-match negative on “free” blocks every search containing that word. Make sure that is what you mean. Exact-match negatives are safer for one-off terms.
• Put them at the right level. Terms that are wrong for the whole business (job seekers, support queries) go on a shared negative list applied to every campaign. Terms that are wrong for one campaign but right for another stay at campaign or ad group level.
• Do not over-prune. Excluding every term that has not converted in a week starves automated bidding of the data it learns from. Judge on enough spend to mean something.

What the report will not show you

Google hides search terms with low volume, so part of your spend will always sit under “other search terms”. Performance Max shows search insights and themes rather than the full row-by-row report, so the routine applies there in a lighter form: review the categories it reports and keep your account-level negative list current. And the report cannot tell you whether a converting term would have converted without the ad. That question belongs to a holdout test, not to this tab.

Fifteen minutes a week is enough. The accounts that waste the most on search are rarely the ones with bad strategy. They are the ones where nobody has read the terms in months.