In short: An audit here is four passes in a fixed order, each producing one piece of evidence, and it stops at the first pass that fails.
Reachability, extraction, resolution, association. Two of them are technical and can be done in an afternoon. Two are editorial and take longer to judge.
The output is one page: the failing pass, the evidence for it, and what the next quarter does. Anything longer is a report about a site rather than a diagnosis of one.
Why It Stops At The First Failure
The passes mirror the sequence a system runs. A page that cannot be fetched is never read, and a page that cannot be read is never quoted, so evidence gathered about a later step is evidence about a hypothetical.
This is the difference between an audit and an inventory. An inventory lists everything that could be improved. An audit finds the one thing preventing everything else, and says so.
The discipline that makes it useful: when a pass fails, stop. Write the finding, write the fix, and do not run the remaining passes until it is closed. Running all four regardless produces a list of twenty recommendations, which is the same as none.
The Four Passes
- 01ReachabilityEvidence: crawler hits in your logs, index coverage in Google and Bing, and the substance of a page present before JavaScript runs.
- 02ExtractionEvidence: a passage taken from the middle of each important page that still answers something on its own.
- 03ResolutionEvidence: an assistant asked about the brand by name returns something accurate, and one canonical name appears everywhere.
- 04AssociationEvidence: the brand appears in answers to category questions, and a named list of who appears instead.
What Each Pass Is Allowed To Conclude
A pass produces one of three verdicts and nothing else: passing, failing with evidence, or not yet assessable because an earlier pass failed.
The third is the one people skip. If reachability failed, the extraction pass has no verdict, because the pages a system would have read were never fetched. Writing "extraction looks fine" from a manual read of the page is an opinion, and it will be quoted back at you.
What You Need Before Starting
Server logs or a log analysis tool covering at least thirty days. Access to Google Search Console and Bing Webmaster Tools. A list of your twenty most commercially important pages. And the question set from module seven, written natively per market.
If the logs are unavailable, say so in the audit rather than substituting a crawler simulation. A simulation tells you what a crawler could do, not what one did, and the difference is the entire point of the first pass.
Scope it to one category and one market first. An audit across five markets at once produces findings that cannot be attributed to any of them, because the retrieval behind each barely overlaps. Run it once properly, then repeat the method per market.
The Output Is One Page
Five things: the failing pass, the evidence, what it costs, the fix, and the date the measurement will be repeated.
Not a deck. A deck invites a discussion about priorities across twenty items, and the whole value of this method is that it has already decided the priority by finding what is blocking.
Key Takeaways
- Four passes in a fixed order, mirroring the sequence a system runs: reachability, extraction, resolution, association.
- Stop at the first failure. Evidence about a later step is evidence about a hypothetical.
- A pass concludes one of three things: passing, failing with evidence, or not assessable because an earlier one failed.
- Logs beat simulations. A crawler simulation tells you what could happen, not what did.
- Scope to one category and one market, then repeat per market. Retrieval barely overlaps between them.
Check yourself
Before you move on
Not scored, not recorded, and not part of the certificate. Both answers are settled by a sentence in this lesson, and the reasoning appears whichever option you pick.
- 01
Reachability fails. What is the correct verdict for the extraction pass?
- 02
Why scope the first audit to one category and one market?