Find the hard side.
It isn’t the scarce side. Most founders aim their first effort at the wrong one and never recover the momentum.
Gate: you can name it, and defend itHow to open a market that doesn’t exist yet — and how to derive your own numbers instead of borrowing someone else’s.
Six procedures. Each ends at a gate you either pass or you don’t. And every number in the book is marked, so you always know whether to trust it.
The problem with the advice
Search for marketplace cold-start advice and you’ll be handed thresholds. Specific enough to read as rigor. Almost always attached to nothing you can check.
A widely shared 2026 guide claims 67% of failed marketplaces die on the supply side, attributed to a16z. a16z’s own published material says close to the opposite.
The same guide’s sample size changed from nine to eleven inside a single page.
That’s the median artifact in this category. Not an unusually bad one.
The method
Six steps. Each ends with a gate you either pass or don’t — so you always know whether you’re allowed to move on, or you’re fooling yourself.
It isn’t the scarce side. Most founders aim their first effort at the wrong one and never recover the momentum.
Gate: you can name it, and defend itThe smallest unit where liquidity is actually achievable — not the market you wish you were in.
Gate: the cell is small enough to fillFrom two constraints — capacity and coverage. Most founders calculate one and never learn the other exists. [the Step 2 model is mine]
Gate: both numbers exist, and you know which assumption is weakestWith a tool that works before the network does. Give the hard side a reason to show up on day one.
Gate: it’s useful with zero counter-partyAnd measure the leak you’re creating while you do it. Manual matching hides a cost; this makes it visible.
Gate: the leak is quantified, not ignoredThere’s a popular 2026 tactic here that will get you sued. The chapter is precise about exactly where the line sits.
Gate: nothing you’d have to retract laterUsing the indicator that moves first — before revenue, before the graph everyone screenshots.
Gate: the leading indicator is trending, not the vanity oneWhy you can trust it
You should never have to guess where a figure came from. So each one carries its provenance, right there in the text.
From your own market, using the procedures in the book. Portable to your cell, not mine.
Every sourced figure points to a reference on the last page. Follow it as far down as you like.
Marked J in the text so you can discount it. There are twelve.
There are twelve J markers. I’d rather you see all twelve than trust an unmarked one. The Step 2 model is mine, and the book says so — including which assumption is most likely to be false in your market, and how to correct for it.
What you get
One purchase. The formats you print and write on are here too — because a field manual you can’t mark up isn’t a field manual.
Who this isn’t for
Every figure in here requires you to go measure something first — and a few of them will tell you things about your cell you’d rather not know. If you want a threshold to borrow, the internet is full of them. This is the other thing.
Get the book
A wrong supply number costs you a quarter you don’t get back. This costs the price of lunch.