Cleaning up your keyword set: which search terms to remove from tracking

Why an oversized keyword set hurts you

Every search term in tracking costs attention. With 20 products at 15 keywords each you look at 300 positions every morning, and the three that really dropped drown in them. Then there is the average: terms you are never found for pull metrics like average position and page-1 share down without anything changing in your business.

The goal is not 'as few keywords as possible' but: every term in the set must answer a question you actually have.

Four kinds of keywords that can go

1. Permanently not found: a term that has reported 'not in the top 120' for three months across every check, while you have not worked on the listing, measures nothing. Either you optimise for it now, in which case it stays, or it goes.
2. Irrelevant: terms your product ranks for but which mean a different variant, size or product type. Such hits bring clicks without purchases and belong not in the set but out of the listing.
3. Reorderings: 'wasserkocher edelstahl 1,7 l' and 'edelstahl wasserkocher 1,7 liter' usually return very similar results on otto.de. Keep the spelling that appears in the search suggestions and drop the others.
4. No demand: a term you derived from a data sheet that neither appears in the autocomplete nor has a meaningful result count is not something buyers type.

What never gets removed

Keep every term that meets one of these conditions: it demonstrably brings revenue (ad statistics or sales figures), it is the main term of the category even if you sit on page 3 there, or it is seasonal and currently out of season. 'Heizlüfter' at position 80 in July is no reason to delete it; in November it is your most important term.

Also keep terms that a competitor serves much better than you do. They are the yardstick for whether an optimisation works.

The quarterly routine in four steps

1. Export: all keywords with current position, best position of the last 90 days and the date of the last listing change.
2. Mark: give each term one of the four removal categories or 'keep'. An hour per 300 keywords is realistic.
3. Decide: for 'permanently not found', ask: do I want to optimise for this? If yes, the term stays and gets a date by which the optimisation should be done.
4. Remove and document: delete the marked terms and keep the list. A term that resurfaces six months later may have been seasonal after all.

What you see after the clean-up

After an honest round a set is often 20 to 40 percent smaller, and the metrics jump: average position improves although no ranking moved. That is not a success but a measurement effect; explain it in the next report, otherwise the improvement looks like an optimisation result. The real gain comes afterwards: a loss on one of the remaining terms stands out immediately, and you react faster.

Frequently asked questions