One site. 12.3× the impressions. Nothing else touched.
Between 2026-03-05 and 2026-08-25 the CRO9 engine ran daily on sxowebsite.com — rewriting the title and intro, measuring, repeating. No redesign, no link building, no ad spend. Every number below is the engine’s own Search Console reading, recorded at the moment of each action.
solid = impressions · dashed = average position (drawn inverted, because a lower position is better)
What actually happened
It got worse first
Average position fell from 55.3 to 81.3 while the engine ran daily. This is in the study because it happened, and because anyone expecting week-one results should know the turn took about six weeks.
The climb
Position 81.3 → 10.0 and impressions 277 → 4,147. The only thing running was the daily engine: rewrite the title and intro, measure, repeat.
A human made manual changes
The site owner applied a few edits picked up from LinkedIn advice. Position went 10.0 → 14.3 and impressions fell about 20%. It is the only stretch in 5.7 months that went backwards, and it is the only stretch where someone edited the site by hand.
Back on the engine
Position recovered 14.8 → 11.9 and impressions returned to 3,900, climbing again at the point this was written.
What this is not
- One site, one observation. There is no control group and no A/B split — we cannot prove what this site would have done untouched.
- Search moves for reasons nobody controls. Algorithm updates and seasonality are part of any five-month window, including this one.
- It is not a promise. We will not quote you a number for your site before we have looked at it — which is exactly why the founders programme starts with an assessment.
What makes it worth reading anyway: the owner made no other changes for 5.7 months, and the one fortnight somebody did edit the site by hand is the one stretch that went backwards.
Book a fit assessment