How an analysis run works

A run is a set of rows in the database rather than a process, which is the reason you can close the tab. Every step it plans is recorded before it is executed, so a run that stops part-way is resumed by the next thing that looks at it — the Continue button, or the scheduler that finishes a run nobody came back to.

What it asks for

The keyword phase goes first. For your own site and for each rival you track — at most 5 of them — it collects the top 1,000 keywords that domain ranks for, ordered by estimated traffic, plus an organic overview and a backlink summary for the domain itself. Your own site also gets a monthly history, so the Dashboard can draw a line rather than a single point.

When that is done the run plans the rest from what it found: up to 10 crawls of your own highest-earning ranking pages, and a set of long-tail searches seeded from the keywords you already rank for. Rivals' pages are never crawled — the audit is about your site.

What you get at the end

• A worklist: the keywords a rival ranks for and you do not, and the ones where you both rank and they are ahead.
• A keyword matrix in Keyword gap, showing who ranks where for each keyword the run measured.
• Technical findings for the pages the run crawled, grouped by how serious each one is.

When a step fails

A failed step fails the run, and the run page carries that step’s own error message rather than a generic one. The exception is the monthly history: nothing downstream needs it, so a run whose history step fails still completes, and the Dashboard says so where the line would have been.

Running again is cheap when nothing has changed: answers are cached for every workspace, so a second run against the same site and market mostly reads what the first one already paid for.