Running a report and reading it

Before

Pressing a run button opens a confirmation first. It shows the endpoints that will be called, the expected number of rows, and the cost estimate.

After the first run of a type, that estimate is based on the previous actual rather than a formula, so it sharpens with use.

Nothing is spent until you confirm.

During

Progress is shown per endpoint rather than as one bar.

Failures are per call, not per run. If one endpoint errors, the run still completes, that section is marked unavailable, and the error is stored with it.

A partial run is still a run

This is worth understanding, because it is unusual. A run that only got half its data is stored, dated and kept — with the missing part visibly missing.

The alternative would be to throw the run away and charge you again, or worse, to show the incomplete result as though it were complete. A gap you can see is more useful than either.

Reading the report

Every report shares one layout across all seven variants, so the second one you read is easier than the first.

Two things to hold on to:

• A NULL is not a zero. “No run has measured that” and “we measured it and found nothing” are different, and the reports keep them different. Do not let one stand in for the other.
• Coverage is always stated. Nothing here is a full crawl. If you quote a figure, quote what it was counted across.

Asking SEB about it

SEB reads stored rows and computed differences only. He never re-fetches and never fills a gap.

He will say “insufficient history” when there is only one run, and he is meant to be visibly less useful early on, because he is. Every claim he makes renders with the run it came from, so you can click through to the rows.

Candidate explanations are labelled as candidates. SEO is full of contested causal claims, and a fluent wrong answer is the failure this product works hardest to avoid.

Export

There are two documents and they are not the same thing — one is the report for a client, the other is the data for you. Pick deliberately.