Does not rank is not the same finding as does not exist

Two findings that look identical in a report and lead to opposite strategies.

“They do not rank for commercial roofing.”

“They do not offer commercial roofing.”

The first is a visibility gap. The second is a portfolio gap. Acting on one when you meant the other wastes a quarter.

What each one can be supported by

Ranking data supports the first. If you have checked their organic positions and their paid keywords and the term does not appear, you can say with confidence that they are not visible for it.

Ranking data cannot support the second. A company can offer a service comprehensively and rank nowhere for it, and that is extremely common. They might be terrible at search. The page might be new. The term might be dominated by aggregators. They might get all their work for that service through referral and never have tried.

Absence from a ranking report is evidence about search visibility. It is not evidence about what a company does.

Only a direct sweep of a competitor’s own site supports “does not exist.” You crawl every page, you inventory the services, and the service is not there. That is a different research action producing a different class of claim.

Why this matters commercially

The two findings point at completely different plays.

Visibility gap. They offer it, they are invisible for it. The opportunity is a marketing one: rank for the term and take demand from a competitor who is not defending it. Timeline is a quarter or two. Risk is moderate, because they can respond by fixing their own visibility, and if they are competent they will.

Portfolio gap. They do not offer it at all. The opportunity is strategic: there is unmet demand in the market, and you can be the one who serves it. Timeline is longer and the investment is real, because you may have to build a capability. But the advantage is durable in a way a ranking advantage is not, because a competitor cannot close it with a content sprint.

Read one as the other and you either spend on content to win a term nobody is competing for because there is no demand, or you build a service line to compete for something your competitor already offers well and simply has not optimised.

The general principle

This is a specific case of a rule worth applying everywhere in competitive research:

Absence of evidence is not evidence of absence.

The rule sounds obvious stated abstractly and gets violated constantly in practice, because research surfaces produce absence as a default output. You search, nothing comes back, and the empty result feels like a finding.

It is not. An empty result is a fact about your search. Whether it is a fact about the world depends entirely on whether your search would have found the thing if it existed.

That question, would this method have detected it, is the one worth asking every time a report says a competitor does not do something.

Where else this bites

AI answer citations. Checking whether a brand is cited in Google AI Overviews or ChatGPT gives you a result for that session, that phrasing, that moment. AI answers are non-deterministic. A brand absent from one response may appear in the next.

So the honest claim is bounded: “we tested six category questions three times each, and the brand appeared in none of the eighteen responses, on this date.” That supports absence for those questions. It does not support “this brand does not appear in AI answers” as a general property.

Ad libraries. Ad transparency records show active creative. They do not show spend, and they do not always show everything. A competitor absent from an ad library on the day you checked may have paused, may be running through a channel the library does not cover, or may have creative that has not propagated. Absence there is weak.

Social and content. Easier, because you can sweep an account directly. If a channel exists and its last post is from 2023, that is a directly observed fact with a date attached. That is strong evidence, and it is strong specifically because you looked at the thing rather than at a report about the thing.

What this looks like in a report

The rule turns into a labelling discipline. Every absence claim carries what it was established by:

No blog, no case studies, no video. Confirmed by direct sweep of their own site, 63 URLs, captured 2026-08-04. Absence established by direct check. Ranking data alone would only support “does not rank,” which is a different finding.

That caveat looks pedantic in isolation. It is the difference between a claim somebody can act on and a claim somebody should not.

The cost of getting this right

It is more expensive. A ranking export covers ten competitors in minutes. Sweeping every page of ten competitor sites and inventorying what is actually offered takes real time.

The alternative is cheaper and produces claims that cannot be relied on, which makes the whole report worth less than its cheapest claim.

The general form: the strength of a claim is set by the weakest evidence supporting it, and mixed evidence quality means the reader has to grade every claim themselves. Most readers will not. They will treat the whole document as uniformly reliable, which means one over-reached absence claim quietly degrades everything around it.

Which is why the discipline is worth the cost, and why it is one of the published rules rather than an internal convention. If we ever report an absence that only a direct check could support, without having run one, that is a failure you should be able to catch us on.