The disavow that did nothing: a useful null result

The disavow that did nothing: a useful null result

Null results are underreported, so this 2022 case matters. A content site convinced an over-optimization problem was throttling it — 38% exact-match across ~600 referring domains — disavowed 210 links carrying commercial anchors over two rounds and waited.

Nine months later: no measurable position change on the 40 tracked terms. Average position held at 14, plus or minus normal noise.

The later finding was that the site had no manual action and no clear algorithmic suppression. Its ceiling was thin content and weak topical depth; the anchor profile, while skewed, was apparently not the binding constraint.

This is instructive precisely because it breaks the reflex that a high exact-match ratio is always the problem. Diagnosis preceded by assumption led to nine wasted months and a slightly weaker link profile.

On one hand, the disavow removing real links with zero downside-or-upside suggests those anchors weren't being counted against the site. On the other, we can't prove the links weren't mildly negative and offset by something else.

Limitation: absence of effect is hard to attribute confidently — the null could hide two cancelling forces.

Open question: how often is 'over-optimization' misdiagnosed when the real ceiling is content quality?