How To Build a PMF Machine
In a world where you can build anything, engineering time is still the most valuable time in your company. A guest post by a16z speedrun investor Marcus Segal.
We're gearing up for New York Tech Week (June 1-7), and we want you to host an event. If you've been wanting to throw a founder dinner, a demo day, a hiring happy hour, or really anything that gets great people in a room together, this is your moment.
Sign up to host here.
This week’s essay is a guest post from a16z speedrun investor Marcus Segal. He draws on his decades of startup experience to offer some time-tested advice on seeking, and keeping, product-market fit.
Emmett Shear, the co-founder of Twitch, visited speedrun founders for a session recently. Someone asked him how he would describe product-market fit. His answer, “How do you describe being in love? You know it when you’re in love.”
Fair enough. But the job of a founder isn’t to sit around waiting to fall in love. The job is to put your best foot forward and to avoid looking for love in all the wrong places. And the combination of a team of engineers cranking out features with AI, a lack of vision for your product and an undisciplined feature list is exactly that—looking for love everywhere all at once with a bad haircut!
When I was at Zynga I was fortunate enough to work closely with Mark Pincus—a founder that had never shipped a game—to build one of the biggest gaming companies on Earth with 300M players. Along the way I’ve co-founded two other startups in B2B and B2C and had the opportunity to support and learn from hundreds of founders on their quests for PMF while mentoring at Y Combinator, Google Launchpad, and now as an investor on the a16z speedrun team.
Across all of these experiences and through massive disruptions in technology three things have never changed:
A CEO’s job is to not run out of money before finding PMF
Engineering person hours are the most important resource in your company
Finding and keeping PMF is more likely to occur with a roadmapping process
It is my hope to share what I have seen work at Zynga and countless other startups in the hopes that all of you can find PMF.
PMF Is a Journey, Not a Destination
When I joined Zynga, none of us really knew how to make great games. What Mark Pincus understood, though, was how to build a system for finding what works—fast.
Our competitors released content once every two weeks, in English. We released five to ten content updates per week, per game, in 26 languages. We didn’t have AI. We generated an enormous number of hypotheses that could be rapidly tested and threw an army of PM’s,engineers, artists and LOTS of QA onto anything that found heat. The work of a 40 or 50 person team at Zynga could now be done with 10 people.
The methodology we used at Zynga wasn’t complicated. The best processes rarely are. Here’s the version I recommend at speedrun—adapted for teams that are building with AI tools to iterate on live products to deliver results faster and better.
Start with the metric, then build the feature.
Before you create a single line of code for a new feature, answer one question: what number are you trying to move? Reach? Retention? Revenue? The specific metric doesn’t matter as much as the discipline of choosing one. Because if you can’t articulate which number a feature is supposed to move, you have no way of knowing whether it worked or was accretive to the product experience. It should be one sentence, “We will launch Product X feature to give Y value to the customer moving Z metric.”
Get all the ideas out, then stack rank them.
This is where most founders stall or face major coordination breakdowns on their teams. All the builders have a running list of features in their heads and they build whatever feels most urgent on any given day. Don’t do that. Start by getting every idea out of your heads and onto a list. Then rank them as a team by asking a simple question: which of these has the highest probability of moving the metric we care about?
T-shirt size the effort.
Vibe coding has broken our intuitions about how long things take to build, particularly across teams.. Pre-AI, experienced engineers got pretty good at estimating effort. Now the volume of code generated across a team of engineers creates new layers of complexity in coordination and design (to name a few).
Things that seemed trivial take forever to polish, and things that seemed hard take twenty minutes and merging everything into a thoughtfully designed UI/UX framework is tougher than ever And let’s not forget We all need to build or rebuild this muscle for the AI era so estimate and track actuals. It’s a new skill so expect a lot of learning along the way.
Power rank by impact-to-effort.
Now you have a list of features, ranked by expected impact, with effort estimates attached. The math practically does itself. High impact, low effort? Do it first. Low impact, high effort? Maybe never.
Roadmap looking one-week-back, one-week-forwards view.
At Zynga, we looked one week back and three weeks forward every week in a team review. What did we build last week? What did we think would happen? What actually happened? The metrics provided the answers. MVP’s of features were tested and scrapped if they did not perform. Course corrections happened to the roadmap to avoid wasting engineering time. Learning compounded quickly and spread across all of our games.
With today’s tools that cadence would be way too slow. Today we look one-week-back and one week forward. The cadence matters. It’s short enough to course-correct fast, but long enough to actually ship meaningful work.
Measure everything.
At every successful startup or scale up I have seen up close the winners are obsessed with metrics. Great ideas can come from any person in your organization.If someone on the team pitches or builds a feature and accurately predicts what it will do they should be lauded. If they are wrong, nobody gets fired. The next time, they will calibrate better. That’s the culture you need. Experience is never wasted if your team learns and gets better.
The Gift
Here’s what I want founders to take away from this: I believe a deliberate roadmapping process is the highest-leverage activity available to you as a founder to increase your chances of finding and keeping PMF. You will maximize your most valuable resource engineering time and grow a killer team in the process.
Enjoy the journey.
Got questions for Marcus? Let us know in the comments. And for more weekly dives into the world of early stage startups, subscribe below.






fantastic read!
I found this insightful!