Where do recommendations go to die?
Between the document and the deployment. A recommendation is agreed, handed over, queued behind other work and quietly forgotten. Months later the page has not changed, the numbers haven’t moved, and the conclusion drawn is that the recommendation did not work.
How does the loop catch it?
A baseline is taken the day a page is added. The suggested changes are shared. Then the live page is checked for them. Until they appear, the page counts as awaiting dev, not as a recommendation that failed.
What happens once the changes are live?
The next reading measures what moved, before → after, against the baseline captured before anyone touched the page. The result is attributed to work that actually shipped.
Why does counting it help?
Because it changes the conversation. “Search isn’t working” becomes “these changes have been waiting three weeks”. That’s a specific problem with an owner, and it can be solved.