Tutorial · Product Management · 18 min

    How to Write a PRD That Engineers Actually Read

    A working PRD template with the sections that matter — problem, users, scope, non-goals, success metrics — and the ones to delete.

    1. Step 01

      Start with the problem, not the feature

      A PRD that opens with "We will build X" has already failed. Open with the problem statement: who is hurting, what they're doing today, and what it costs them (time, money, errors, churn).

      If you can't write the problem in three sentences without using the word "solution," you're not ready to write the PRD yet — go do five more customer interviews.

      Template

      # Problem
      <Who> is trying to <job-to-be-done>, but <current friction>.
      This costs us <measurable impact> per <period>.
      Evidence: <tickets, interviews, analytics>.
    2. Step 02

      Define users and their context

      List the primary user, secondary users, and the operator (the person who supports them when things break). Each gets one paragraph: who they are, what device/channel they're on, what they know, what they don't.

      For FinTech: always name the compliance/ops user. They're who you'll be debugging with at 2am.

    3. Step 03

      Scope: in, out, and explicitly not

      Three columns, not two. The "explicitly not" column is what saves your sprint. Examples:

      • In scope: P2P send between two wallets, same currency.
      • Out of scope (this release): Bulk send, scheduled send.
      • Explicitly not: Cross-border send. Will require a separate PRD with treasury sign-off.
    4. Step 04

      User stories and acceptance criteria

      One story per behavior, Gherkin-style acceptance criteria. Keep them testable.

      Story: As a sender, I can send money to another wallet using their phone number.
      
      AC1: Given a valid recipient phone, when I confirm, the transfer completes in <5s.
      AC2: Given an invalid recipient, the UI shows "Number not registered" within 1s.
      AC3: Given insufficient balance, the confirm button is disabled with reason shown.
      AC4: Every successful send writes one ledger entry per leg (debit + credit).
    5. Step 05

      Success metrics — leading and lagging

      Pick one headline metric and 2–3 supporting metrics. Define them with the SQL you'd run on day 30.

      • Headline: % of new wallets that complete a P2P send within 7 days of activation. Target: 35%.
      • Supporting: Send success rate, median time-to-confirm, support tickets per 1000 sends.
    6. Step 06

      Risks, dependencies, open questions

      Three lists. Each item has an owner and a due date. "Open questions" is where you put the things you don't know — be honest, it builds trust with engineering.

    7. Step 07

      What to delete from your PRD

      Wireframes go in Figma. API contracts go in the design doc. Sprint plans go in Jira. Marketing copy goes in the launch doc. The PRD answers what and why, not how.