Operations and decision log
A public record of how Know Plain keeps volatile facts current, reviews product quality, and connects material decisions to evidence.
Weekly volatile-fact review
Owner: editorial operations. Escalation: engineering when a canonical fact or calculator changes.
- 1. Check ACA legislation, CMS announcements, IRS guidance, and SSA updates represented by volatile records in the source ledger.
- 2. Record the check date even when the value is unchanged; update the canonical fact and every dependent test when it changes.
- 3. Run unit, content-token, calculator-evidence, and production-build verification.
- 4. Add a visible correction when a published output was materially wrong, then document the change in the annual changelog.
Monthly product-quality review
Owners: product and engineering. Review accessibility failures, broken links, zero-result categories, checkup completion, calculator boundaries, community reports, corrections, performance, and privacy-policy conformance. New journey changes require a hypothesis, primary metric, guardrail, and rollback condition before release.
Product funnel dashboard
Instrumentation is deployed. Baselines remain explicitly pending until a complete 14-day production window exists; no sample or fabricated counts are shown.
| Stage | Rate | Initial target | Baseline |
|---|---|---|---|
| Acquisition | Checkup Viewed ÷ Homepage pageviews | 20% | Pending production data |
| Activation | Checkup Completed ÷ Checkup Viewed | 65% | Pending production data |
| Value | Recommended Action Opened ÷ Checkup Completed | 35% | Pending production data |
| Retention | Result Saved Local + Checkup Result Printed + Checkup Lead Captured ÷ Checkup Completed | 15% | Pending production data |
Baseline rule: Do not promote targets to commitments or rank popular shortcuts until the full window has real production data.
Reversible release controls
Material interaction changes ship with an owner, hypothesis, metric, guardrails, and rollback condition.
localToolScenarios
Enabled- Owner:
- Product and engineering
- Hypothesis:
- Explicit device-only save controls reduce repeated data entry without increasing privacy confusion.
- Primary metric:
- Result saved locally
- Rollback:
- Disable if save/restore errors exceed 1% or users reasonably mistake local storage for cloud backup.
checkupEstimateLabels
Enabled- Owner:
- Product and editorial
- Hypothesis:
- Explicit rough-estimate labels increase completion without making provisional results look final.
- Primary metric:
- Checkup completion
- Rollback:
- Disable if completion falls materially or usability review finds the labels create more confusion than honesty.
Decision log
2026-07-14
Keep detailed checkup inputs in the browser by default.
The checkup does not need server storage to calculate a result, and balances, income, and debts create avoidable privacy risk.
Inspect supporting policy2026-07-14
Use ranges and threshold maps instead of confidence-sounding single forecasts.
Withdrawal-rate methods answer different questions, while ACA and Medicare costs change discontinuously at legal thresholds.
Inspect supporting policy2026-07-14
Require moderation before community submissions become public.
Retirement questions are high-stakes and can attract scams, personalized securities advice, and unsafe identity disclosure.
Inspect supporting policy2026-07-14
Deny analytics properties by default and monitor failures only by coarse class.
Raw search, error, URL, and form context can contain health details, balances, email addresses, or tax information.
Inspect supporting policy