What a request costs you
Usage split (requests / user / month)
—margin %
—users underwater
—break-even usage
Seat vs usage vs hybrid, side by side
Usage-based bills cost plus a fixed markup so margin is constant; seat-based is flat, so it wins on light users and loses on heavy ones; a hybrid caps the downside with an included allowance.
| Cohort | Cost | Seat revenue | Margin |
|---|
Flat pricing on a usage-based cost is a bet on your user distribution
Because the LLM under your product bills by the token, a flat per-seat price is really a wager that the average user will consume less than the seat covers. When usage is heavily skewed — a long tail of light users and a handful of power users — that wager pays off, because the many light accounts each throw off pure margin while only the few heavy ones lose money. But if your usage distribution is even, or your best customers are also your heaviest, the same flat plan quietly turns your most engaged users into your biggest losses. The break-even usage is the seat price divided by your blended cost per request; every account above it is underwater. This calculator splits your base into light and heavy cohorts, prices each under a flat seat and under a cost-plus usage model, and shows where a hybrid plan with an included allowance would protect the margin. Pin down your real cost per request first on the cost per 1,000 requests calculator, check the raw model prices on the model comparison, and pressure-test a single price on the AI wrapper margin calculator.
Frequently asked questions
Should an AI product charge per seat or per usage?
It depends on how evenly your users consume tokens. A flat per-seat price is simple to sell and gives customers a predictable bill, but because the underlying LLM cost is usage-based, a per-seat plan quietly overcharges light users and loses money on heavy ones. If most users are light and only a few are heavy, per-seat can be very profitable — the light majority subsidizes the heavy minority. If usage is spread evenly or skewed toward power users, per-seat pricing bleeds, and usage-based or a hybrid with an included allowance plus overage protects the margin. This calculator shows the margin under a flat seat price given a light/heavy usage split so you can see which way your particular user base leans.
What is a power user and why do they lose money?
A power user is someone whose token consumption pushes their raw API cost above the flat price you charge them. Under seat-based pricing everyone pays the same, so a user who runs ten times the average number of requests costs you ten times as much while paying the same seat fee — every extra request past the break-even is a direct loss. The break-even usage is simply the seat price divided by your blended cost per request: consume more than that and the account is underwater. A healthy seat-based plan needs the average user to sit well below break-even so the loss-making tail is covered by the profitable majority.
How does a hybrid included-allowance plan work?
A hybrid plan charges a flat base fee that includes a generous but bounded allowance of usage, then bills overage per unit above it. It keeps the simplicity buyers like for the common case while capping your downside on power users, who pay for what they consume past the allowance. The art is setting the allowance high enough that the vast majority never see an overage line — so the plan still feels flat — while low enough that genuine power users cross it and cover their own cost. Set the allowance near the break-even usage this calculator reports, and you get predictable bills for most and protected margin on the tail.