Founders often ask us to build the product and the pricing page in the same breath, as if the revenue model is a detail to settle once there are users. It is the opposite. The model decides what the product has to be good at, and reversing that choice after launch touches billing, data, sales, and every customer contract at once.
The model is an architecture decision, not a marketing one
Per-seat billing means the product has to be worth rolling out to a whole team, and the permission model has to be real from the first release. Usage billing means metering has to be accurate to the cent on day one, because the meter is the invoice. A free tier means the free path has to be cheap enough to serve at scale without a human in the loop. Each model pushes the build in a different direction — picking one late means reworking parts that are already load-bearing.
Match the model to how the buyer already budgets
The cleanest pricing bills the way the customer already thinks about the cost. Sales and operations teams buy seats. Infrastructure and data buyers buy usage. Finance signs an annual contract. When the model cuts against how the buyer's budget is structured, every deal picks up friction that no amount of discounting removes.
This is worth testing in real conversations before it is baked in. Two or three calls with target customers about how they would expect to be charged will surface a mismatch far more cheaply than the first renewal cycle will.
Price on value, meter on something the customer controls
The metered unit should rise when the customer gets more value from the product, and it should never produce a surprise. Billing on a number the customer cannot see or predict — background API calls, internal compute, retries they did not trigger — breaks trust the first time an invoice jumps, and that is usually the month they were about to expand.
Write down what would make you change it
Decide in advance the signal that the model is wrong: gross margin below a set line, sales cycles stretching past a limit, churn concentrated in one tier, or expansion revenue that never materialises. Put a number on each.
Without that, a broken model gets defended for a year, because changing pricing feels like admitting the first call was a mistake. Named thresholds turn it into a scheduled decision instead — the same discipline we apply to architecture, where the point is to make the expensive change a plan rather than a fire drill.