Fusion Trace / Individual and bulk workspace
Follow the evidence.
Keep the result.
Use your Quarry API key for individual or bulk lookups, inspect returned information and reopen saved results. Keep evidence and uncertainty visible at every step.
For real lookups, buy credits in Quarry, accept its use policy and create a scoped API key. Individual and bulk requests use the same Quarry balance; no separate Trace key is required.
Generated conceptual imagery. No customer data or verified result is shown.
The customer workflow
From a lookup to a saved result.
01
Access
Buy Quarry credits, accept its use policy and create a scoped API key. Use that key for individual and bulk requests.
02
Submit and poll
Start a single lookup or upload a batch. Poll the existing job instead of submitting it again.
03
Review
Inspect the returned sources and confidence. A returned contact is not a guarantee that you have the right person.
04
Save and export
Results are saved to your Quarry workspace. Reopen owned jobs and lists and export saved results without a new lookup.
Everyone else sells
hit rate. We publish measured accuracy.
A 70% hit rate only means a phone number came back. It does not mean the number reaches the right person. We rank for right-party contact — and every number we return carries its evidence: which independent sources agreed, how stale the record is, and what its confidence turned out to be worth when checked against a government record.
- 13%
- right-party rate
- n=200
- 32%
- number found at all
- the ceiling
- 631
- graded predictions
- 64 positive
- 57s
- median trace
- free tier
Measured against a state voter roll of 9.2 million registrants, with that roll switched off during the trace. The full curve, including the band that claims 60% and delivers 17%.
The guest demo contains fictional people and never runs a real lookup. Real requests require Quarry credits, accepted use policy and a scoped API key.
Existing clients: legacy workspaceA number nobody ever checked.
Every provider in this market reports a confidence score. Almost none of them have ever compared those scores to a set of known-correct answers. The score is produced by machinery that was never told whether it was right.
Right party, not any party
Candidates are anchored to the subject's metro, so a same-name stranger three states away is rejected instead of returned. Attach a paid source and address history anchors it further.
Independent families
Two people-search brokers reselling one upstream feed is one source, not two. We count independent families and show you the count.
Current, not merely known
“Is this line still live?” is a time-to-event question. Line type and last-reported date feed the answer on the free path; carrier and live-line status need a paid lookup, and the dossier says which it had.
You don't assert the truth. You earn it.
There is a calibration harness, it runs against a labelled gold set, and its reliability curve, sample size and per-source scores are published on the certificate — including the part where it currently refuses to certify anything, because the positive class is too small. Until it does, every confidence on this platform renders as uncalibrated — never as a guess dressed up as a measurement.
See the current numbers →Owner name + address, farmed from public sources. Ours, not rented. Planner estimate.
Self-hosted. No per-lookup fee on the free tier.
If we could not verify it, the field is empty and says why.
Measured right-party rate, sample size, reliability curve.
Run one and judge the output.
Try fictional examples without an account, or use your Quarry key for real requests.