Download

Product teams

Run the PRD before engineering builds it.

Your spec has been reviewed, commented on and approved. Has anyone actually tried to use the feature it describes? Subjourney sends simulated personas through your PRD, so the gaps surface in planning, not in QA.

Run your first PRD
  • Product managers who want to know the spec holds up before they hand it off.
  • Product designers looking for the states and flows nobody has designed yet.
  • Engineering leads who'd rather answer questions at planning than mid-sprint.
  • Founders writing their own specs, with no one else to catch the gaps.

The cheapest bug is the one in the spec.

You know the pattern. Engineering asks a question two days into the sprint. Design fills a gap with a guess. QA finds the edge case nobody wrote down. Every one of those started as a missing sentence.

And now that AI drafts half your specs and coding agents build from them, those missing sentences ship faster than ever.

What a run catches

  • Missing

    The unhandled state.

    The returning user with an expired trial. The PRD covers new and paid, and the persona falls between them.

  • Ambiguous

    The silent assumption.

    “Users invite their team” assumes they have permission to. Your admin persona doesn't.

  • Contradictory

    The conflicting requirement.

    Section 2 says the setting is per workspace; section 5 says per user. The persona can't satisfy both.

  • Dead end

    The missing exit.

    A flow with a way in and no way back out.

Run the docs you already write.

  • PRDs and one-pagers, including AI-drafted ones.
  • Feature specs and acceptance criteria.
  • User stories and epics, run together as one journey.
  • User flows and diagrams from your design files.
  • Pricing, plan and permissions docs that the feature depends on.

Review tells you how it reads. A run tells you if it works.

  • Spec review

    Today
    Comments on wording and structure
    With Subjourney
    Evidence of where a user gets stuck, and why

  • Edge cases

    Today
    Whoever thinks of them in the meeting
    With Subjourney
    Personas drawn from the users your spec describes

  • Handoff

    Today
    Questions arrive mid-sprint
    With Subjourney
    Questions answered before kickoff

  • AI-drafted specs

    Today
    Read it through and hope
    With Subjourney
    Run it and see

  • Scope changes

    Today
    Old sections quietly go stale
    With Subjourney
    Re-run shows what the change broke

Between “draft” and “ready for build”.

  1. 1Draft the PRD, by hand or with AI.
  2. 2Run it against the personas who'll use the feature.
  3. 3Fix the paragraphs that stalled them.
  4. 4Hand engineering, or your coding agent, a spec that's already been walked.
  5. 5Re-run when scope changes, so the doc stays true.

In practice

A PM runs a new billing-change PRD with three personas: a team admin, a solo user, and an admin mid-way through a downgrade. The first two make it through. The third stalls at “Confirm plan change”: the spec never says what happens to seats over the new limit. One paragraph added, one re-run, and the question engineering would have asked on day three is answered before kickoff.

A run, step by step

One PRD, three personas, two findings.

  1. 01Import.

    The PM brings three docs: the billing-change PRD, the pricing page copy, and the seat-management spec.

  2. 02Model.

    Subjourney builds the journey.

    Change plan
    Open billing
    Choose plan
    Review changes
    Confirm plan changeFlagged
    Receipt

    While building, it flags that Confirm plan change has no failure path.

  3. 03Run.

    Three simulated personas set off. Choose one to see what they want and what gets in their way.

    Solo userSubscriber

    Switching to annual billing.

    Motivation
    Lock in the annual discount now the tool is part of every week.
    Frustration
    Being charged twice when a plan changes mid-cycle.
    Need
    A plain note on what happens to the month already paid.
    • Age29
    • Income$58,000
    • EducationBachelors Degree
    • EmploymentEarly Career
    • Communication styleCasual
    • Social engagementActive Contributor
    • Risk appetiteRisk-Tolerant
    • Eco awarenessEco-Conscious
    • Work-life balanceFlexible
    • Brand loyaltyActive Loyalty
    • Price sensitivityValue Seeker
    • Tech adoptionEarly Adopter
    • Tech literacyAdvanced

  4. 04Findings.

    MissingStalled at: Confirm plan change

    Admin downgrading, 14 seats, new limit 10

    The PRD never says what happens to seats above the new limit. Expected after Section 4, paragraph 3.

    Open in sourceSend to agent

  5. 05Fix and re-run.

    The PM adds a paragraph on over-limit seats and aligns the pricing copy. Same three personas, re-run: all three reach the outcome.

    Two paragraphs changed. The question engineering would have asked on day three, answered before kickoff.

Personas to start with

Five runs that catch most spec gaps.

  • The happy-path user.If they stall, nothing else matters yet.
  • The returning user.Old data, old settings, a plan that's since changed.
  • The one without permission.A member trying what only an admin can do.
  • The one halfway through.Mid-migration, mid-checkout, mid-trial.
  • The one who leaves and comes back.What does your spec say happens to their progress?

Questions

How is this different from asking an AI to review my PRD?

A review comments on the doc. A run uses it. The persona tries to complete a task with only what your spec says, and every stall points to a paragraph.

My specs are short. Is there enough to run?

Yes. A thin spec just stalls sooner, and each stall shows you what to write next.

Does this add a step to our process?

It replaces one. Run it before the review meeting, and the meeting becomes about decisions, not hunting for gaps.

Can findings go to our coding agent?

Yes. Send findings to the agent you build with, then re-run once it's patched.

Hit play on your PRD.

Bring the spec you're about to hand off.

Download Subjourney