An anonymized adaptation: what the counts show, and what they cannot
An anonymized and lightly normalized adaptation of a public audit. The firm is a long-established contract engineering and technology advisory firm with three service lines. Page and URL counts are rounded, service lines are described generically, and no name, product, competitor, or founding date is given. The central findings are as the audit reported them. Its identifier in the toolkit is CASE-CE-01.
The counts show a correct name and no inclusion on three service lines
| Observation | Count |
|---|---|
By-name answers the audit judged correct |
All of them; the firm’s own site was cited in most |
Buyer-style answers on its three service lines that named the firm |
0 of 95 (45, 25, and 25 answers) |
Other providers named per service line |
At least one in about a third to three quarters of the answers |
Indexable pages |
About a dozen |
Service pages that return 404 |
Three |
URLs found that redirect |
Most of them |
Structured data, or a file listing the site for language models |
Neither |
The competitor counts are derived from rounded percentages and the providers are not named. The audit’s questions were a shared set, not written from a Finder Specification for this firm, so only the three service lines are reviewed as the firm’s own. The scorecard scores Entity Recognition, Finder Specification Coverage on those slices, and Cross-Engine Consistency. It leaves six dimensions unscored and computes no overall score. Three technical checks are known, which is below the six required, so Technical Legibility is not scored.
The findings keep the observation, the hypothesis, and the check apart
Observation. The firm was named in 0 of 95 answers on its three service lines, and other providers were named in about a third to three quarters of them. Hypothesis. Public evidence connecting each service line to the firm’s capabilities may be too thin or too poorly linked for engines to use. The counts do not show which. Would disconfirm. Providers with equally thin public evidence for the same problem being included consistently, or the firm staying absent after it publishes evidence in the finder’s words. Then thin evidence is not the explanation. Experiment. For each line, open the sources cited where other providers appeared, record the source types, and compare them with what the firm publishes for the same problem.
Observation. Three service pages return 404, most found URLs redirect, and no structured data exists. Hypothesis. These may keep crawlers from treating the service pages as the firm’s canonical evidence. Nothing here shows they did. Experiment. Repair them first and change no content. If nothing moves after the engines have had time to refresh, the repairs probably were not the limit, and movement after a later content change is more likely the content’s.
Three interventions start with the cheap repair
IV-1 (technical, days 1 to 30): restore or redirect the three failing service pages and remove redirect chains. IV-2 (content, days 31 to 60): one page per service line that states the problem in finders' words and the firm’s method, with a delivered example the client agreed to share. IV-3 (evidence, days 61 to 90): independent corroboration for the first line, such as a client reference, a talk, or a publication, each real and citable. Beside each is the result that would show it failed. All data is in the toolkit download under examples/contract-engineering-firm/.