PM craft: decision logs, unsell emails, and automating user research

Executive overview

Most PMs drift toward internal work — alignment, stakeholders, process — and lose contact with the raw material of their job: customers and decisions. Kevin Yien argues that product sense is just good decision-making under uncertainty, and it improves only through deliberate reps.

The core practices he shares — keeping a decision log, drawing tight constraints for your team, unselling job candidates, and automating a stream of customer conversations — are all ways to stay close to reality while scaling your judgment.

The PM's job is to convert the potential energy of a team into maximum realized value — with minimum loss.

How to enter product management

• Don't go straight into PM. Start as an engineer, designer, or salesperson first.
• Those roles are the smallest, fastest loop between building and customer feedback.
• Sales is underrated: the best salespeople are the best listeners and translators of customer problems.
• PM is not a fallback from those roles — it's where you double down if you discover you're potentially world-class at it.
• Companies like Stripe, Figma, and Twilio delayed PMs because founders were the customer. That only works when you're building for yourself.

Writing as a core PM skill

• Writing is clarity at scale — the primary vehicle for creating alignment across a team.
• PMs who only write internal docs miss half the job. You should be able to write in the voice of the person you're serving.
• If you can't sell or support your own product in words, you shouldn't be building it.
• To become a better writer, consume writing that compels — Paul Graham essays, specific voices with crisp thinking.
• Vary sentence cadence deliberately: monotonous sentence length causes the reader to tune out.
• Recommended: Several Short Sentences About Writing — teaches tight sentences as a foundation.

Drawing the perimeter: the PM's role with engineering and design

• The PM's primary job with engineers and designers is to define constraints, not solutions.
• Constraints are what let engineers and designers go as creative as possible within a meaningful boundary.
• Key constraints to set: who exactly is the customer, what job they're trying to do, which platforms matter, and what trade-offs are non-negotiable (e.g. speed over consistency).
• Avoid hard swim lanes. The murky overlaps between PM, design, and engineering produce better work.
• PMs must stay obsessed with the final deliverable — not just the kickoff.
• Example: spending a week tuning an animation's timing (in milliseconds) for a restaurant point-of-sale, bringing in actual bartenders and servers to test prototypes.

Getting better feedback on documents

• Don't send a doc to Slack and hope for comments. Schedule a "silent read" meeting instead.
• In the meeting: no talking, everyone reads and comments in real time, author responds in real time.
• The compressed feedback loop — comment, response, follow-up — is far faster than async threading.
• This is moving slower to move faster: higher latency tolerance in scheduling, dramatically lower latency in the actual feedback cycle.

The decision log

• Product sense is the ability to make good decisions with insufficient data. It only improves through deliberate reps.
• Keep a running log: record decisions you make or observe, write your rationale, set a calendar reminder to review outcomes.
• Extend this beyond your own role: what would you do if you were responsible for another team's roadmap? Another company's product launch? Write it down. Check it in six or twelve months.
• Format: a single rolling document (or note), tag decisions with #decision, use command-F to search.
• Start small: once a week, find something interesting, place your bet, set a reminder.
• Caveat: this is a complement to building, not a substitute. Reading Hacker News alone won't make you a better PM.

Hiring: the unsell email

• Unsell email: at offer stage, send a candid email listing the things most likely to trigger the candidate's fears or doubts.
• Through the interview process, collect the anxieties the candidate hints at but doesn't say explicitly.
• The email should have no more than eight bullet points and be direct — e.g. if you're a Series A startup with weekend expectations, say that plainly.
• If a candidate reads the full email and is still equally excited, you have a high-conviction hire.
• If they hesitate, it's better to know now than after six months.
• When Kevin first used this, 30% of candidates dropped at offer stage. Recruiters objected. He sent the email anyway.
• Don't send it at the start — you need the full interview process to know what their real fears are.
• After sending: most candidates will want to talk through one item. Make yourself available at whatever time works for them.

Automating customer research

• PMs need direct exposure to raw customer conversations — not processed reports, not sales summaries.
• If you're not allowed to talk to customers, that's a structural problem that must be fixed first.
• Two tools for a steady stream of B2B customer conversations:
  • UserintroHQ (userintro.com): user testing focused on B2B; you set ICP criteria, they source candidates, interviews land on your calendar automatically.
  • Gong + Slack + Zapier: set keyword alerts in Gong (e.g. product terms, competitor names), route matches to a Slack channel, use Zapier to pull the customer's email and drop your Calendly link into an automated outreach sequence.
• The result: a continuous, automated pipeline of relevant customer conversations — no manual sourcing.
• Credit for the Gong workflow: Beth Hills, PM at Mutiny.
• The trap to avoid: telling yourself you already know the customer well enough to stop talking to them. Their lives change. Your mental model must stay current.

On failure and identity

• Being laid off — especially early in a career — can feel like evidence you're not cut out for the role.
• The critical reframe: separate "I'm not good at this" from "this company didn't need what I offer right now" or "my way of working didn't fit this environment."
• Many strong PMs and designers fail in one company and flourish in another. The environment matters enormously.
• Performance conversations feel hard because they're heard as "you are bad." They're more useful reframed as: does your working style fit this machine, or should you find a different one?
• Use failure moments to distinguish what's truly part of your identity from what's within your control to change.

AI and what's coming

• The real shift isn't the current capabilities of LLMs — it's the mental model of the next generation.
• Kevin's daughter encountered Midjourney V1 and asked "did I draw that?" — treating the model as a crayon, not a tool.
• Children growing up with generative AI as a crayon will have entirely different intuitions about what a good product looks like.
• What they'll expect in 20 years is not possible to predict from our current vantage point.

Books and inputs

• The Courage to Be Disliked — Eulerian psychology; focus only on what you control. Kevin re-reads it annually.
• The Paper Menagerie — sci-fi/fantasy short stories; recommended for lateral thinking.
• Autobiographies and memoirs: the fastest way to spend 50 hours with someone you want to learn from.
• Paul Graham essays for crisp, clear writing.