What a client's first hour on DataTap actually looks like — measured, not described · September 2026
Before our first client sent a byte, we ran the onboarding runbook cold on a throwaway tenant, timed every step, and kept what we found. Here is the whole cycle.
| Provision the tenant + first admin | 333 ms | empty profile, four-eyes on |
| Dry-run validate a 631-row export | 19 ms | full report, zero writes |
| Ingest it for real | 77 ms | 631 accepted, 2 markets derived from the data |
| Upload 25 KYC profiles + rebuild the graph | 867 ms | entity resolution over the upload |
| First T+1 scan | 350 ms | every detector, against real baselines |
| Generate the PoV report | 19 ms | WORM-stored, print-ready |
| Offboard (purge 698 rows) | 116 ms | the DPA's deletion promise, as code |
Machine time for a complete onboarding cycle: under two seconds. The 90-day proof of value is long because of humans and data, never because of the platform.
Three bugs — all the same class: demo fixture data leaking into a real client tenant.
Every fix carries a regression test. The published changelog lists them in the open. We would rather you read this here than discover it on your own data.
A benign export produced zero alerts on its first scan. That is the correct first impression: a surveillance program that manufactures findings from clean data is noise, and noise is the thing compliance teams stop reading. The detectors fire on the planted ground truth they're tested against — and on nothing else. You can watch that live in any demo workspace with the red-team button.