Work · Product design and engineering
Product Health Scorecard
An assessment for founders deciding where to investigate a live product. I owned the product decisions and implementation, using AI-assisted development. Its email system checks consent when a message is sent, so a withdrawn permission also affects messages already queued.
- Stage
- Shipped build
- Evidence
- Working build
- Last updated
- 25 August 2026
- For
- Founders whose product is already live and not growing as expected.
- My role
- Sole owner of the product decisions, engineering, copy direction and operations. Development was AI-assisted. No other human contributors.
- Latest decision
- Enforce email consent when each message is sent, not only when it is queued.
The problem
Founders with a live product that isn't growing usually know something is wrong before they know where to look. The Scorecard asks about the product and the business around it and suggests where an investigation could begin. The result is a starting point based on the answers, not a diagnosis.
The important decision
The assessment promises that follow-up emails only go to people who agreed to them. I decided that promise had to be enforced by the system every time an email is sent. Each message is checked against current consent when it is claimed for sending, so withdrawing permission suppresses messages that were already scheduled.
How the system works
A 20-question assessment across ten dimensions routes to a personalised result and a report email. It is built with Next.js, React and TypeScript, with Supabase Postgres and Resend called through their REST APIs. A transactional outbox handles retries, de-duplication and per-message unsubscribe tokens, and a scheduled job dispatches due email.
What I checked
I verified the production flow end to end in August 2026 and exercised the consent, suppression, de-duplication and unsubscribe paths against a local stack, alongside an automated test suite.
What remains open
The scoring thresholds and routing are untested with real founders, and the field test has not run. I'm not claiming any completion, accuracy or conversion figure.
What happened
The full flow — session, submission, personalised result and report email — was verified end to end in production in August 2026. No founder has completed it as part of a field test yet, so there are no user-behaviour results.
Next question
When founders complete it, does the suggested starting point match where their investigation actually leads?
Artifacts
- Building note: The Promise Is Infrastructure25 August 2026
Look closer