Capture the Evidence

Capture the Evidence is the second phase of the Data Suitability recipe: "Fix a baseline, record the certified evidence, and retain it for governance review." It turns the AI Use Case built in Define the Use Case into an attestation you can hand on. This page describes its three steps and the certification that sits between them.

Baseline

Fix the referent drift is measured against.

The step opens with an explanation of why a baseline exists: "Drift since baseline" is meaningless without an explicit referent. Capturing one records today's quality scores, profile identifiers, and row counts for every bound container. Every attestation in this certification cycle reports movement from that referent to its evidence capture, and the AI Use Case compares that evidence with the current state afterwards.

Click Capture Baseline. A card confirms Baseline captured, with the capture date and how many containers were snapshotted. Once a baseline exists the button reads Start New Baseline, with a warning beside it: starting a new baseline replaces the active referent and returns the record to Draft. Existing evidence remains unchanged, but a Manager or Admin must certify the new cycle. See Start a New Baseline.

Capturing needs at least one bound container. Next requires a baseline.

Evidence

Capture the certified, point-in-time evidence record.

The step first lists what the attestation records: the current quality dimension scores and their explanations per bound container; movement from baseline to attestation in scores, row counts, re-profiling, and re-scoring; the check set with its pass and fail split; field-level lineage into the bound assets; and a masking and authorization summary. It carries governance evidence only, never a complete source record or preview row. The comparison from attestation to current state is kept on the AI Use Case and refreshed on demand.

Without a baseline the step warns that the server will refuse to attest and sends you back a step.

Certify before recording evidence

Certification is the governance approval; the attestation is the immutable evidence of that approved state. Recording evidence first would preserve a Draft or In Review decision in the artifact, so the step shows the record's certification badge and explains How approval works:

Submit for Review changes the record from Draft to In Review. It does not certify the model.
Certify records the approval and enables evidence generation. It is a separate job from authoring: it takes a Manager with the Editor team permission on every bound datastore, or an Admin.

The step then shows only the transition buttons you are allowed to use: Submit for Review, Send back to Draft, Certify, Revoke, and their counterparts for an expired or revoked record. A certifier looking at a record In Review also sees Certification valid through (optional), a date to leave blank when the approval has no automatic expiry; the scheduled expiry check uses this date. When the record is In Review and you cannot certify, the step says a Manager with Editor permission on every bound datastore, or an Admin, must open it and select Certify, and reminds you that Qualytics does not send an automatic approval notification, so share the page with the approver. See Certify an AI Use Case and, for the lifecycle, AI Use Cases.

Generate the attestation

Generate Attestation is enabled once the record is Certified and a baseline exists. It creates the attestation and shows a result card, Evidence record #N captured, with the containers covered, the open anomalies, the check total and how many are failing, and the attestation it supersedes, if any. The button then reads Generate a New Attestation, with the note that regenerating supersedes the current record and nothing is overwritten.

Generation is refused, with a message, when the record is not certified, when no baseline exists, or when the bound containers changed after the baseline was captured. In the last case, capture a new baseline and certify the new cycle. See Troubleshooting.

Next requires an attestation.

Complete

Download the artifact and hand it on.

The step confirms Evidence is on the record: the use case now carries an immutable Data Suitability evidence record, including its certification context and the evidence captured at that point in time. Under Hand it to your auditor, three actions are offered:

Download Evidence JSON, the complete point-in-time evidence record.
Download HTML Report, a readable rendering of the same record for an audit file or governance review.
Open the AI Use Case, which opens the record's page, where the Attestations tab keeps the history and the live comparison.

Start Another returns to the record list to begin a new use case. Nothing is deleted: the finished use case keeps its record and attestations. See Export an Attestation.