How we measure reputation
A Reputation Audit walks a fixed list of eleven review and profile platforms, asks each whether your brand is there, and reads back the rating, review count and response rate it can see.
Depth: judgment draws
Depth reruns only the model's judgment. Platform figures come from one fetch per run, and a second draw would replay identical bytes, so we neither resample nor charge for them: five draws cost about 1.6 times one. A rung shows the best grade that many draws can earn; your report is stamped with what it measured.
| Judgment draws | Grade | Cost | What it buys |
|---|---|---|---|
| 1 Default | T1SnapshotOne read. Rerun to confirm it. | ×1 | |
| - | T2Directional | - | Not offered. The ladder steps from T1 straight to T4. |
| - | T3Consensus | - | Not offered. The ladder steps from T1 straight to T4. |
| 2 | T4Established*Rerun, and at least 4 in 5 reads matched. | ×1.15 | |
| 3 | T4Established*Rerun, and at least 4 in 5 reads matched. | ×1.3 | |
| 5 | T4Established*Rerun, and at least 4 in 5 reads matched. | ×1.6 | |
| - | T5Verified | - | Not offered. |
Why it stops at T4. Agreement is not correctness: five wrong reads can still agree.
* Our estimate, not yet validated by real runs.
Disagreement is a finding: the model has no settled read of your brand, and the report shows which problems every read named and which only some did. Agreement is not correctness, and the odds favour it: the judgment runs at very low temperature and picks from a short list of mapped problems and fixes, so a disagreement is loud. Every sampled run records how many options there were. Failed draws stay in the denominator: two of three answered, never a unanimous two.
Breadth: platforms reached
Reach over a roster we wrote, fetched all at once, so there is no order to read a curve from. Breadth is how much of the roster answered, weighted as the coverage score weights it: Google Business Profile and Trustpilot 0.4 each, TripAdvisor 0.2, every other platform 0.1.
| Platforms | Grade | Cost | What it buys |
|---|---|---|---|
| - | T1 | - | Not on the scale. Narrow is the lowest breadth grade. |
| 2 | T2Narrow*New finds had not slowed yet. | ×1 | Google Business and Trustpilot. |
| 3 | T3Partial*Still finding new things at the end. | ×1 | Google Business, Trustpilot and TripAdvisor. |
| 4 | T3Partial*Still finding new things at the end. | ×1 | The three majors plus Facebook. |
| 11 Default | T4Mostly saturated*New finds had nearly stopped. | ×1 | Every platform we read. |
| - | T5Saturated | - | Not offered. |
Why it stops at T4. Eleven platforms is our whole list, not the whole internet.
* Our estimate, not yet validated by real runs.
The run stamps its own reach, as plainly as "3 of 11 platforms reached". We tested grading breadth as a discovery curve over the probe order: it called 58% of every audit ever run Saturated, including audits that found one profile in ten probed. A platform that errored is unfetched, not absent: it counts against breadth and leaves the score's denominator.
What the report claims
| Number | Type | What it can support |
|---|---|---|
| Rating, review count, response rate | Counted | What the fetch returned, never a claim that the platform is right about you. |
| Star average, weighted coverage, combined score (60/40) | Derived | Never more certain than the counts beneath them. |
| Named problems and first fix | Interpreted | Graded by agreement, because they come from a short closed list. |
| Written summary | Interpreted | A single read by construction: two paragraphs saying the same thing in different words are not a disagreement. |
Fidelity: mostly a stand-in
Outside Google Business Profile, the platform surface grades T1Proxy. We would rather say so than let a clean number imply we loaded the page.
| Surface | How we read it |
|---|---|
| Google Business Profile | As a listing, from a business-data lookup: rating, review count, claimed status and recent review text. The closest to the real surface, and weighted heaviest. The only platform whose review text we fetch. |
| Trustpilot, TripAdvisor, Yelp, Facebook, G2, Capterra, Glassdoor, Indeed, Booking.com, Expedia | From a site-scoped Google search: the rating in Google's rich snippet, a cached summary of the page. It can be hours or days behind, and a regional host carries its own count. |
| Booking.com, Expedia | Rated out of ten. Normalised to five for the average, shown with the listing's own figure. |
| Profile matching | A guess that has been wrong: one bank matched a sibling's profile with eight reviews while its real listing held about 4,500. Every card carries a "Wrong URL?" link, and a pinned profile overrides matching. |
Glassdoor and Indeed measure what employees think, not customers, yet sit at equal weight in one star average. Glassdoor is the second most common extra platform we find, so reweighting would move scores customers have already seen; it is a live decision, not an oversight. If employer reputation is not what you measure, switch those two off in the template.
What we do not claim
- That we read the review pages themselves outside Google Business Profile.
- That the platforms are right about you.
- A confidence percentage on a model judgment, at any depth.
- That a platform we could not reach is one you are missing from.
- That our list of eleven is the whole of your reputation.
- More than one draw where a number rests on one. The report says so.
