Engine Trace
Show the regulator the working, including everything that came back empty.
Every screening keeps a record of all 117 sources it queried, what each returned, what was dropped as a wrong-entity match, and which parts of the pipeline did not complete.
Engine Trace is the record of what a screening actually did. It lists every one of the 117 sources that was queried with its outcome, the articles that were dropped as a different person or company and why, and any part of the pipeline that degraded. Nothing is discarded silently, so a clean result can be defended.
- Sources on the record
- 117
- Per-source outcome
- Returned, empty, skipped or failed
- Dropped matches
- Listed with the reason
- Broken lanes
- Flagged, never blank
How it works
A search that keeps its own receipts
- 1
Every source query is logged with an outcome
The engine records a status for each of the 117 sources it tried, not just the ones that returned something. The result page shows the count of sources with results out of the total, and the source panel lists each one.
- 2
Wrong-entity matches are kept, marked as dropped
Adverse media and news that mention a same-name person or company are removed from the findings by an identity gate, but they are not thrown away. They are stored as dropped items with the reason and shown in their own list, so you can check the gate did the right thing.
- 3
The org-chart engine reports its phases
For company subjects, the collapsible Engine Trace panel shows what the leadership engine did: discovery of the corporate domain and team pages, deep scrape counts, harvest queries run and returned, extraction by bucket, wrong-entity and low-trust drops, registry merges and the final kept count.
- 4
Each lane says whether it finished
Ownership tracing, corporate structure and the other lanes each record ok or failed with a short reason. The UI shows an amber notice for a failed lane, so an empty section reads as nothing found only when the lane genuinely completed.
- 5
The same record goes into the PDF
The report carries a Basis of this report section, a Coverage section listing what was searched, an Excluded matches section for the dropped items and a Limitations section for anything that did not complete.
What you see in the app
Three panels on every result
Sources panel
A headline count such as "31/117 returned results", then every source with its status.
- success
- the source answered and results were kept
- no_results
- the source answered and had nothing on this subject
- skipped
- the source did not apply, for example a company-only registry on a person
- missing_key
- the source needs credentials that are not configured
- rate_limited, blocked, unreachable
- the upstream refused or timed out
- cached_stale
- a stored copy was used because the upstream was unavailable
Dropped, wrong entity
A collapsible list under adverse media. Each item shows the title, the outlet, the snippet, the reason the gate removed it and the confidence. Demoted items that turned out to be neutral news appear in the news section instead.
Same-name-different-person is the most common false positive in adverse media screening. This list is how you prove you handled it.
Engine Trace panel
Collapsed by default on company results. Opens into phases: discovery, deep scrape, LinkedIn people, harvest, extract, filter, merge and final, each with counts. Lane notices sit beside it: "Corporate structure" and "Ownership graph" show ok, or a reason if they degraded.
Why it matters to a regulator
A clean report is only worth something if you can see what was searched
The question in an examination is rarely "did you find something". It is "how do you know there was nothing to find". A screening that returns no hits and no record of coverage cannot answer that. Engine Trace turns "no hits" into "these 117 sources were queried on this date, these 13 official lists at these versions, these were unreachable, these matches were excluded for these reasons".
The design rule inside the engine is that a lane which degrades must say so. An empty officer list and a broken officer engine used to look the same. They no longer do, on screen or in the PDF.
Pair it with the AI council for a reproducible score and Reports for the document you file.
Frequently asked questions
What does Engine Trace show that a normal report does not?
The negative space. Which sources were queried and returned nothing, which were skipped or unreachable, which articles were removed as a different entity and why, and which parts of the pipeline degraded. A normal report shows hits; Engine Trace shows the search.
Are dropped articles really kept?
Yes. The identity gate stores dropped and demoted items with the reason and a confidence value. They appear in the result under a dropped, wrong entity list and in the PDF under Excluded matches, so a reviewer can disagree with the gate if they want to.
What happens when a source is down?
Its status is recorded as unreachable, rate limited or blocked and the screening continues. Official sanctions feeds are also snapshotted twice a day, so a list can still be checked from the stored copy and the status says cached.
Can I export the trace?
The source outcomes, exclusions and limitations are rendered into the PDF report on every plan. The full screening result, including the trace fields, is also returned as JSON by the API.
Run a screening and read the trace.
Plans start at $49 a month for 50 screenings. Every plan includes the AI council, the ownership graph, Engine Trace and PDF reports.