Tight ad groups still win, even in the age of broad match

Every couple of years the advice swings back to “consolidate everything, let the machine sort it out.” Fewer ad groups, broader match, one asset pool, more data per campaign. The argument is real: modern bidding needs volume, and slicing an account into forty thin ad groups starves every one of them.

But consolidation and looseness are different things, and they keep getting confused. You can run a small number of ad groups and still keep each one tight. What you cannot do is put unrelated intents in the same bucket and expect the ad, the landing page, or the negative list to be right for all of them.

What broad match actually changed

Broad match changed which queries reach a keyword. It did not change the fact that one ad group serves one ad and one landing page. If “emergency boiler repair” and “boiler service plan” sit together, the ad has to hedge between an urgent problem and a planned purchase, and the hedged version loses to both of the specific ones.

The cost shows up as a click-through gap first and a conversion gap second, which is why it often goes unnoticed: the campaign still spends, still reports conversions, and the underperformance is invisible without a comparison.

Group by what the searcher wants next, not by what the keywords have in common. Two phrases sharing a noun is not a shared intent.

Group by intent, not by theme

Thematic grouping is seductive because it is easy to automate: sort by shared token, ship. Intent grouping takes a judgement call — is this person comparing, buying, or panicking? — but that judgement is exactly what determines the right headline and the right page.

A practical test: write the headline you would use for the group. If you cannot write one headline that is genuinely the best answer for every query you expect in that group, the group is doing two jobs.

What the search terms report keeps proving

Read the search terms report as a structure document rather than a bill. Every query that arrived in the wrong group is telling you either that a negative is missing or that an intent you have not built for is out there with money behind it.

The second case is the valuable one. Repeated queries you never planned for are the cheapest new-ad-group research available, and they arrive already priced. Most accounts only ever mine the report for waste and never for opportunity.

What to check in the first fortnight

In the first two weeks, look at the split between queries that match the group’s intent and queries that merely match its vocabulary. If more than a small minority are vocabulary matches, tighten before you optimise anything else — bid adjustments on a mixed group just average two different truths.

After that, consolidation is fair game. Merge groups that turned out to share an intent, and let the data pool. Consolidate deliberately, from tight to loose, rather than starting loose and hoping the machine reconstructs a structure you never wrote down.