Raw answer captures
Every figure traces back to the stored answer that produced it. This is the feature.
Every run's answer is stored verbatim, with its provenance, and is reachable from every metric it contributed to. Click a figure, get the runs; click a run, read exactly what the engine said.
What provenance records
| Field | What it pins down |
|---|---|
| Method | Grounded API, or SERP vendor, and which one |
| Engine and version | So a step-change can be attributed to a model update |
| Locale and geo parameters | What location the question was asked from |
| Language | Which phrasing, in which language |
| Timestamp | When, to the second |
search_triggered | Whether the engine actually performed a web search on this run — it frequently does not |
Why this is the trust feature
The sharpest criticism of this whole tool category is that three tools given the same prompts produce three different answers, and none of them will show you why. The test independent reviewers now apply is simple: can every aggregate metric be traced to stored, inspectable observations? If it cannot, the number is a claim.
No metric is auditable without access to the observations behind it.
Retention
Captures are the provenance spine, so they are not aggressively expired — but they are not kept forever either. Answer payloads are erased on a twelve-month window or on request, while the result row survives as a tombstone so that a link to an erased capture answers “this was deleted” rather than “this never existed”. See Retention and erasure.
See it against your own client list
A working demo runs your prompts, in your market, on live engines — not a sandbox with seeded data. Bring one client brand and three competitors.