Make the founder the support channel so every bug, friction point and feature request lands in your own DMs; overwhelming, but it is how you learn the core problem.
When it applies: early stage, founder-built product with X audience
Watch for customers bending the tool to a job it wasn't built for; then pivot, rebrand and buy a new domain around that job instead of force-selling the original.
When it applies: product with churn or stalled growth
Filter, don't comply: check each request against what you're building, keep some value reserved for higher tiers, and refuse requests that make you like everyone else. Revid deliberately refuses scene-by-scene manual editing and generates a full video out of the box.
When it applies: post-PMF products with a clear differentiator
He handles support himself on each new product instead of relying on dashboards, following PG's 'do things that don't scale'; he says a week of manual work teaches more than a month of analytics.
Add a customer chat on every page, watch for recurring requests, move the patterns onto a public feedback board users vote on, and build from that instead of assumptions.
When it applies: Post-launch SaaS with active users
He uses ShipFast to start every new product and only adds what he personally needs (e.g. landing-page components built for ZenVoice fed back in), rather than building every user request.
Add a customer chat widget, write a doc for each repeated question, and link it in-product at the exact step users get stuck. Polish docs like product UI.
Seven months in, harsh feedback hurt; he stopped defending and asked why it failed. The issues were specific and fixable: AI prompts not tuned for real use cases, onboarding built for himself not a first-time user, and a bloated dashboard ('300 tools and 500 fancy buttons'). Fix the 20% of problems causing 80% of churn.
When it applies: Early product with churn and critical feedback
Talk to churned users; when one says they stay on a competitor only for feature X (auto-DMs), build that rather than spending time on imperfect-but-working systems like SEO caching bugs. Separate non-ICP complaints (e.g. student discount requests) from signal.
After joining a product, change its core to match what you think people want, run it with a small group for weeks or months, learn what makes it sticky and stand out, then launch broadly.
When it applies: taking over or relaunching a product