Real answers, written out in full. Ask Pricing Brain above for the version that fits your numbers, your metric and your customer base.
Do not flip the whole base at once. Pick the usage metric that already tracks the value customers get, then price it so a typical account lands within 10 percent of what it pays today. Launch the new model on new business first, grandfather existing contracts to their renewal date, and give every account a side-by-side view of old price versus new before they are asked to move. Revenue leaks in migrations come from surprise, not from the model.
It lowers month-to-month certainty and raises expansion. You get that certainty back with a committed platform fee plus usage above it, so a floor is contracted and the upside floats. Most companies that report volatility after switching priced pure consumption with no commitment. Track net revenue retention and the share of revenue under commitment rather than looking only at monthly recurring revenue, which stops meaning the same thing.
Price the work completed, not the tokens burned. Find the unit of output the customer would otherwise pay a person to produce, such as a resolved ticket, a qualified lead, a reconciled invoice, and price a margin over your compute cost on that unit. Then set a floor commitment so margin holds when usage is thin, and a cap or tier break so a heavy month cannot invert your gross margin.
The right value metric grows as the customer succeeds, is easy for them to predict before they sign, and is cheap for you to measure and invoice. Test candidates against those three at once. Seats fail the first when your product replaces work rather than seating people at it. If no single metric passes, use a hybrid: one metric for the base fee and a second for expansion.
Usually as three layers. A platform fee covers access and support, a usage layer charges for the automation actually run, and an overage rate applies above the committed volume. The key decision is the billable event. Charging per workflow run rewards inefficiency, and charging per successful outcome aligns better but needs a definition of success both sides accept in writing before signature.
Look for near-universal win rates, discount requests that almost never appear, buyers who approve without escalation, and accounts whose usage vastly exceeds what they pay. Any one of those on its own is weak evidence, and three together mean you are leaving money on the table. Confirm with structured willingness-to-pay research rather than a gut raise, so you know where the ceiling sits before you test it.
Set tier breaks where customer behaviour already changes, not at round numbers. Chart usage of your value metric across the base, find the natural clusters, and put a boundary between them so most accounts sit comfortably inside a tier rather than fighting a ceiling. Each tier should have one clear reason to leave it. If you cannot name that reason in a sentence, the tier is packaging noise.
Size the increase from research, not from what feels bearable, and stage anything above 20 percent over two renewals. Announce it at least 60 days out, name what changed on your side, keep the current price available for a longer commitment, and give the largest accounts a call before the email. Churn after an increase tracks how it was communicated more closely than how large it was.
