Gate trials behind a card so only committed users sign up; a two-person team can't support 500 free signups a week who never pay. Use it for B2B or configuration-heavy tools; skip it for viral consumer apps where volume matters.
When it applies: B2B, setup-heavy tools, small teams; not viral consumer apps
Auto-assign each new user to an experiment variant (headline, monthly vs annual-first pricing display, onboarding step count/order, CTA copy), measure one metric at a time in PostHog or Amplitude, ship the winner. Revid uses exports as the success metric.
Small teams can't absorb freemium support; 200 users at $9 eats your time while 20 users at $99 makes the same revenue. Price so only people who need it pay; consumer apps are the exception.
When it applies: teams of 1-3 selling business software
Give the free tier one job: if it drives word-of-mouth and future buyers, count it as CAC; if it serves people who never pay, it's charity. AI products have real per-use costs, which is why some limit free usage.
When it applies: freemium products, especially AI with marginal costs