Private beta policy
Sunday Lab privacy
Sunday Lab is an invitation-only NFL forecasting lab operated by Measured Studios. It uses practice credits only. There are no payments, deposits, withdrawals, or external account connections.
Identity and access
The public marketing site does not create an account or grant access. When the private application is activated, Google sign-in will be handled by Clerk. PostgreSQL will own accounts, invitations, pool membership, and authorization.
Beta access requests
During the temporary marketing launch, Netlify Forms processes and stores the submitted email in plaintext with group size, adult acknowledgement, and this policy version. The form collects no free text, files, or campaign attribution. Netlify is not an account, invitation, or application-data authority.
Request retention and migration
Submission-email notifications and routine CSV exports are disabled. Requests are reviewed at least weekly and permanently deleted from Netlify no later than 30 days after submission. At the Google Cloud cutover, selected leads will be asked to resubmit through the encrypted first-party intake. Netlify history will not be imported automatically.
Interaction evidence
Forecasts, decisions, positions, and settlements remain in their authoritative product tables. Separate telemetry records only interaction actions such as opening a brief or citation, beginning a post-mortem, or saving a lesson. It excludes email, provider subject, rationale, probability, line, stake, and free-form text.
Retention and pseudonymization
Raw interaction events are retained for 90 days and then reduced to aggregate counts. The planned account pseudonymization workflow removes external identity linkage, replaces profile labels, redacts member-authored free text, revokes access, and preserves deidentified numeric scorecard history with a restricted audit receipt. Backup retention is capped at 35 days, and erasure receipts must be replayed after a restoration.
Cited analysis
Game briefs retain structured claims, source metadata, citations, model and prompt versions, cutoffs, and output hashes. Full publisher articles are not stored. Source links receive non-identifying campaign attribution only when rendered.