Every crawler produces a list sorted by its own idea of severity. Severity is a property of the issue type — a blocked page is worse than a skipped heading level — but it says nothing about the page. The same missing title on a page with ten thousand impressions a month and on a page with none are not the same job, and a severity sort puts them side by side.
The alternative takes one export from Search Console and an afternoon. It produces a list whose top item is worth doing this week, which a severity list almost never does.
The steps
- Export impressions per page for the last 28 daysSearch Console, Performance, Pages tab, export. One row per URL with impressions and clicks. This is the only external data the method needs.
- Join the audit findings to it by URLEach finding now carries the impressions of the page it sits on. Findings on pages Search Console has never shown carry zero, and most of your list is about to become zero.
- Group findings into causesSame issue type, same URL pattern, same template: one cause. Sum the impressions across the pages in the group. How to do the grouping is its own guide; the short version is that most sites have a few dozen causes behind a few thousand findings.
- Rank causes by impressions times severityMultiply the group's impressions by a small severity weight — three for an error, two for a warning, one for a notice — so that a warning on a high-traffic template still outranks a critical on a forgotten one. The exact weights matter less than having them.
- Do the top three, then measureNot the top thirty. Three causes, fixed properly, with a control set frozen when each one ships, so that in four weeks you know which of the three did anything. Then re-rank; the list has changed.
Why impressions and not clicks or rankings
Impressions are the measure of how often Google already considers a page a candidate. A page with impressions and a defect is a page where the defect is costing something today. Clicks measure the outcome of the title and snippet as well as the position, so they are the right metric for judging a title fix but the wrong one for ranking the list. Rankings for a chosen keyword measure one query and miss the long tail where most impressions live.
This is the ranking Rankelle uses for its impact score, and what we do not do includes inventing a traffic estimate for pages Search Console has no data on. Zero is the honest number.
Questions people ask
What about pages that have no impressions because of the issue?
A noindex or a robots block produces zero impressions precisely because of the defect, so the method underweights it. Treat indexability findings as a separate, short list reviewed by hand: they are few, and each one is either a mistake or a decision.
Is 28 days the right window?
It is the shortest window that covers four weekly cycles, which matters because impressions swing by weekday. Longer is fine for ranking; it is only for measuring a fix afterwards that the window has to start when the fix landed.
Try it on your own site, free
The whole audit of every website you add, no card and no expiry. Pay only when you want the proof.