Quick Start
The fastest useful outcome is not a green dashboard. It is one representative experience with attributable evidence, explicit manual boundaries, and owned barriers.
1. Create a project
Name the product or service being evaluated. A project contains experiences, barriers, components, conformance scopes, and records; setup itself does not remain a permanent navigation destination.
2. Capture an experience
Open Experiences and choose one path:
- Enter a public
https://URL for hosted discovery. - Use the browser recorder for an authenticated or multi-step journey.
- Import or run a CLI definition when an engineering-owned contract already exists.
Choose a complete process that matters to users. Confirm every constituent step before approving the draft scope.
3. Set representative contexts
Begin with contexts that reflect actual user and release risk: browser, viewport, zoom, locale, direction, theme, reduced motion, and authentication or data state where relevant. Avoid multiplying combinations without a reason.
4. Run the baseline
The evaluation stores rendered and accessibility evidence, rule-contract
versions, and ACT-style outcomes. Review failed and cantTell results rather
than interpreting an incomplete automated result as conformance.
5. Resolve barriers
Open Barriers to consolidate repeated occurrences, assign an owner, capture remediation, approve time-limited exceptions where policy allows, and verify a fix through a new execution. Closed history is retained so recurrence remains visible.
6. Build conformance deliberately
In Conformance, define the product version, standard and level, technologies, representative sample, complete processes, exclusions, evaluators, and dates. Criterion decisions may combine mechanical evidence, guided manual evaluation, user testing, exceptions, and verification.
Publication is blocked while required scope, evidence, decisions, evaluators, or approvals are incomplete. Reshot does not certify accessibility or guarantee acceptance of a report.
7. Add delivery policy when calibrated
Repository and CI integration is optional. Start advisory. Enforce only
approved, mechanically supported new failures after the team has measured false
blocks. Human review, cantTell, untested, worker failure, unavailable
services, and active-flow overage never become false accessibility failures.
Next, read the Evaluation Methodology, Conformance Workflow, or the private-pilot Record Overview. Add delivery policy only after calibration.

