Unit economics for PMs
Reason about what a feature earns and costs so product bets connect to the business.
A feature's value isn't what it does. It's what it earns minus what it costs to deliver — per user, at scale.
The number that decides whether the feature survives
Most feature debates happen in the language of usage: engagement, adoption, delight. Those are real, but they're upstream of the question the business is actually asking — does this unit of product make money, and does the answer improve as we grow?
You don't need a finance degree for this. You need a rough, defensible read on two things: what one more user of this feature is worth over their lifetime, and what one more user costs you to serve. When those two lines diverge in the wrong direction, no amount of engagement saves the feature.
Building a back-of-envelope read
- 01
Name the unit
A user, an account, a transaction, a support ticket. Everything downstream is 'per one of these.' Pick the one the business actually scales on.
- 02
Estimate what it earns
Revenue attributable to the unit over its life — expansion and retention included, not just the first payment.
- 03
Estimate what it costs to serve
Infra, support, payments, and any human in the loop. The costs that recur per unit, not the one-time build.
- 04
Check the direction at scale
As units grow, does margin per unit improve, hold, or erode? A feature that gets more expensive per user as you grow is a slow leak.
- 05
Locate the lever
If the number is bad, is it a price problem, a cost problem, or a retention problem? Each points to a different owner and fix.
Where the unit math bites
A generous free storage tier
Nothing directly — free users don't pay.
Real storage and bandwidth that scales linearly with sign-ups.
Only worth it if free users convert or refer enough to cover their carry. Dropbox's model worked because they did; many copycats bled out because they didn't.
An AI-powered summary feature
Marginal — it's a nice-to-have inside an existing plan.
A per-call inference bill that grows with exactly the users who love it most.
Your best users become your most expensive. The feature can be a hit in usage and a loss in margin at the same time.
White-glove onboarding
Higher activation and retention on the accounts that get it.
Human hours that don't shrink as you add customers.
Sound for high-ACV accounts, ruinous for self-serve. The same feature is a great bet or a terrible one depending on the unit it's attached to.
Pick a feature you're proud of. What does one more heavy user of it cost you — and does that cost fall or rise as the product grows?
You don't have to model the P&L. You have to know, roughly, whether the thing you're shipping earns more than it costs per user — and which way that gap moves at scale.