Search term waste: from report to negative keywords in four steps

A keyword is the phrase you bid on; a search term is what the user actually typed. Bid on “office chair” in broad match and you will also show for “office chair repair” or “second-hand office chair”. That difference is where understanding your spend begins.

The search terms report exposes those matches. In any account that is not reviewed regularly, it is normal to find a meaningful share of the budget draining into terms that never convert.

There are three main sources. First, broad and phrase match stretching the intent: people looking for information, jobs or free solutions click but do not convert. Second, competitor brands: the user typed a specific brand, clicked your ad and bounced. Third, out-of-area or out-of-scope searches: a city you do not serve, a model you do not sell.

These three patterns repeat regardless of industry. The good news is that because they repeat, they can be blocked in advance.

Step one: pull the search terms report for the last 30 or 90 days. Step two: filter for terms with zero conversions that spent above a threshold; set the threshold relative to your average cost per conversion. If the average is $15, non-converting terms above $20 are the first candidates.

Step three: group the terms by pattern: “free”, “how to”, competitor names, cities. Step four: add the pattern as a phrase-match negative rather than each term individually, so new variations of the same pattern are blocked too. Done by hand, those four steps can take an hour.

In Tecrube the same analysis is one request: “group search terms from the last 30 days that spent over $20 without converting, and propose a negative list”. The data is read from the live account, the model does the grouping and writes the reasoning, and the result arrives as a preview.

The preview shows each negative term, its match type and the campaign or shared list it will go to. Remove the rows you disagree with, then approve. Writing to the account happens only after approval. The same flow works for Microsoft Ads; carrying the list you approved for Google over to Microsoft is a second approval.

Take a boiler repair business. When the last 30 days are pulled, most of the non-converting spend clusters into three patterns: “boiler prices” (purchase intent, not service), “boiler error codes” (informational intent from people fixing it themselves) and the names of districts the business does not serve.

The assistant's suggestion is three rows: the “prices” pattern as a phrase negative, the “error code” pattern as a phrase negative, the district names as exact negatives. In the preview you notice “boiler fault” is on the list too; that term is actually service intent, so you remove it. You approve the rest.

A week later, the same question: “what happened to the non-converting share after the negatives?” If the ratio fell, the first cycle is complete; if new patterns appear, they go on the list. Because the cycle can be repeated in chat, the analysis stops being a one-off project.

Instead of scattering negatives campaign by campaign, use shared lists at account level: general waste patterns, competitor brands and out-of-area searches in separate lists. When a new campaign launches, the lists are attached and the first week's waste is cut from the start.

Make the analysis a weekly routine. In Tecrube the 75 free credits are enough to start; afterwards the Pro plan sustains the routine for one account. One caution when adding negatives: blocking a high-intent term that merely has not converted yet is also a cost, so never approve a preview without reading it.