Every SEO report has gaps. An account was never connected, a data source has a delay, a metric is not available for a particular page type, or a tool could not reach a page. What a report does with those gaps decides whether the client can trust the rest of it.
The three kinds of “nothing”
When a number is missing from a report, it can mean three different things, and they should look different:
- A measured zero. The metric was collected, and it was zero. Zero clicks from a query is a finding.
- Not collected. The source was not connected, or the period was before connection. The honest label is “not connected”.
- Unavailable. The source could not supply the figure for this period or page. The label is “unavailable”.
Some reporting setups collapse the second and third into the first. The chart then shows a drop to zero that looks like a ranking collapse, a client can read it as one, and time is spent explaining a problem that never happened. The reverse is just as bad: an estimate filled in quietly looks like a measurement, and the report is eventually caught out when someone checks it.
Why this is an SEO issue, not just a reporting one
SEO reporting is how clients judge whether the work is paying off, and the data comes from places with real limitations: crawl data from a particular date, search data with its own definitions, analytics that depend on correct tagging, third-party tools that estimate. A report that is explicit about what it knows and does not know can be defended line by line. One that papers over gaps can be undone by a single wrong figure, and the damage spreads to numbers that were right.
Published reports should stay published
A related discipline is that a report, once sent to a client, should not change. If figures are recalculated later, perhaps because a data source was corrected, the old report should still show what the client was told at the time, with the correction recorded in a new one. Otherwise nobody can say what was reported in March.
Documentation follows the same rule
We applied this to SEO Ops itself. Some of its integrations could connect to a data source before they could pull data from it, and until the data sync shipped the guides said exactly that: the connection works, and the figures show as unavailable. They were rewritten when it shipped, not before. Describing a feature as finished before it is finished is the documentation version of a made-up number.
AI-drafted analysis needs a human before the client sees it
The same principle applies to text that is generated. Drafting commentary, summaries and recommendations with AI saves time, and it also produces confident sentences that nobody has checked against the data. Google’s guidance on generative AI content is blunt on the point: it asks site owners to check and review all AI-generated content for accuracy and trustworthiness before publishing
. Its spam policies warn that using generative AI to produce many pages without adding value may violate the policy on scaled content abuse.
The workable approach is a clear status: an AI draft is labelled as a draft, a person reviews and edits it, and only then is it approved and published. That is how AI drafting works in SEO Ops, and the product’s own wording for it is “AI drafts. People decide.”
Where the gaps come from
It helps to know the usual sources of missing data, because each has a different honest description:
- Not connected. The data source was never linked, or access was withdrawn. Nothing was collected, and that is the whole story.
- Permissions or quota. The source is connected, but the account cannot return that figure, or a limit was reached.
- Delay. The figure is not yet complete for the period. A partial number presented as final is the commonest accidental error.
- Definition changes. The tool changed what it measures, so before and after are not comparable.
- Before measurement existed. Tracking began on a particular date, and earlier periods are unavailable by definition.
- Failure to retrieve. Something went wrong on the day. The figure exists, but this report does not have it.
These are not all the same, and a client who is told which one applies can judge how much it matters. An estimate is acceptable when it is labelled as one, with its method stated, and never when it is placed in a column that otherwise contains measurements.
How to word a gap
Missing data is easier to accept when the report explains it plainly and says what happens next. A few examples of the tone to aim for:
- “Search Console is not connected for this site, so organic clicks are not shown in this report.”
- “Rankings for these keywords could not be retrieved on 12 September. The figure will appear in the next report.”
- “Organic sessions before 1 March are unavailable, because tracking was not in place until then.”
Each one says what is missing, why, and what the reader should do about it, which is usually nothing.
Annotate what changed
Charts that cross a change in measurement need a note. If tracking was fixed in the middle of a period, a tool was swapped, a definition changed, or a site migrated, say so on the chart, because a step change in the line is otherwise read as a real change in performance.
Use one source of truth
Figures that are retyped between systems drift. A report built from stored data, with the period and source attached to each figure, can be regenerated and checked. One assembled by hand cannot, and the first time someone asks where a number came from, nobody can say.
A short test for any SEO report
- Can you tell, for every number, where it came from and for what period?
- Does a missing figure say so, rather than showing zero?
- Is anything an estimate? Is it labelled as one?
- Has a person read every sentence that was written by a machine?