An empty field or dash does not mean that a check was completed and no errors were found. The result may not be available yet, the URL may not have been included in the crawl, or some responses may still be unknown. Before drawing a conclusion, find out what was actually checked.
A zero also needs to be interpreted in the context of the metric. A measured value of zero and no measurement are different states. Don’t replace unknown data with zeros just to make a table look tidy.
For an audit, check completion and the URL in question
Check the status and date of the result. A pending or in-progress check has not produced a final result yet. If a previous report is visible beside it, that report is for an earlier date.
Then look for the original URL. The daily automated audit covers up to 100 pages. If the page where you fixed an error has no entry in the new crawl, the absence of a finding does not confirm that the issue was fixed.
Example: yesterday, /catalog/chair had an empty page description, but today the URL is not on the list of errors. Several things may have happened: the description was fixed; the page was not checked; or the check has not finished. First establish what happened with that URL, then close the task. The process for verifying a finding can help.
For search and AI data, keep the conditions of the result
If a position was not found, the result applies to the selected query, region, and results obtained. It does not prove that the domain is absent from search everywhere. Check the date and what was included in the comparison.
A manually added competitor may have no measured metrics. Don’t invent an average position based on a dash. Internal visibility is an estimate based on tracked queries, not market share or visits; interpret a zero alongside the positions found.
For AI, having no observations before setup does not mean zero mentions. If a company is not named in the responses obtained, the conclusion applies only to those questions, models, and the period. Models that did not respond should not be counted as negative responses.
You can review the purpose of the sections in “Site Monitoring”. Assess your own site using the data actually obtained, keeping values unknown wherever there is no answer.
In a working table, describe what the dashes mean instead of using the same dash for everything
The examples below show how to record different cases:
| What we see | What to record in the task | Next action |
|---|---|---|
| The audit is still in progress | The new result is not available yet | Check the completed result |
| /catalog/chair is not in the crawl | This page was not checked | Check the specific URL separately |
| No position found for one query | No position was found in this result | Check the query, region, and date |
| One AI response is missing | Only responses from the models that responded are known | Check the status of the missing response |
| No finding for the checked URL | This issue was not found at this URL | Record the date and method of the check |
The last row is useful, but it does not confirm all the other properties of the page. A successful check of the page description does not verify that an employee receives a submitted request.
You could write to the client: “There’s no result for this page yet because it wasn’t included in the crawl; Anna will check the URL on 6 October.” This communicates both the limitation and the follow-up without showing a misleading zero.
Before sending a summary, find every zero and empty field
For each value, ask: Was the metric measured? What does it mean? Which URL or set of queries does it refer to? If there is no answer, add an explanation or keep the status as “not checked.”
Pay particular attention when transferring data to a table or presentation: an empty cell can become a zero, and its label can get lost. Agree on notation with colleagues—for example, a dash means there is no usable value, while zero is a measured numerical result. This is a document convention, not a change to the service’s calculations.
Once the data is available, update the relevant row with the new date. The guide to source freshness may help with comparisons. An unknown result first requires a check; assign a fix once a specific issue has been found.