How to move a meter without a mutiny

A meter is not a price you set once and forget. It's a relationship you run. The number on the pricing page is a single moment; the meter behind it is a thing the customer lives with every day, watching it tick, forming opinions about whether it's fair. Which means every choice about how it behaves, and every change you ever make to it, is not a config update — it's a trust event. The companies that forget this don't lose because their new economics were wrong. They lose because of how they changed the deal.

Two of the most public pricing fights of the last two years prove the point by what didn’t happen. When Cursor reworked its $20 plan in mid-2025 so the same money bought roughly half the usage, the backlash was fierce enough to force a public apology and refunds within weeks. When GitHub Copilot re-based its meter to bill on tokens consumed, heavy users watched their bills jump and said so, loudly. But notice what the customers didn’t demand in either case: they didn’t flee the product, and they didn’t insist on the old flat model forever. The pricing model survived both fights intact. What got punished was the manner — the abruptness, the opacity, the sense of a deal changed from underneath them. The lesson sits at the center of this whole chapter: how you move the meter matters more than where you move it to.

The four levers that decide how the meter feels#

Before you ever change a thing, four settings determine whether your meter feels fair on an ordinary Tuesday or feels like a trap waiting to spring. Get these right and you bank trust you can spend later; get them wrong and you’re starting every renewal in a hole.

Exhaustion behavior — what happens when the meter hits zero. Three options: a hard stop, a throttle, or automatic overage. Across the roughly two hundred plan tiers in the AI Credit Index, where it’s stated at all the most common answer is a hard stop — but a meaningful share quietly default you into auto-overage: keep working, get billed, find out at month-end. That is a budget the buyer doesn’t control, and it’s the single fastest route to a furious renewal call. Default to a soft landing — pause or throttle with a clear prompt — and make paid overage something the customer switches on, not something that happens to them.

The top-up premium — what the refill costs versus the bundle. When both numbers are visible, the Index shows the median top-up runs about 12% more per credit than the plan it tops up, with the worst cases charging three to four times more — and the steepest premiums land, predictably, on the cheapest entry tiers, exactly where the buyer has the least room to move and feels the squeeze most. A modest, transparent premium is fair: you’re carrying the spike risk, and that’s worth something. But gouging the customer in their busiest week — which is precisely when they’re cornered — trades one good month for a reputation you don’t get back.

The peg — what a credit is worth. Dollar-pegging a credit at a clean cent makes the unit legible, and buyers like legible. But it quietly boxes you in: if a credit is always a cent, the only lever left to raise the price is to change how many credits an action burns. The friendly peg is an IOU — you’ve promised a stable unit price, and the only way to collect is to move the thing behind it. Compute-pegging is more honest about the variability but harder to sell. Pick the trade you can defend, and go in knowing what the cent commits you to.

Expiry versus rollover. Included credits almost always expire at the end of the cycle. Pair that forfeiture with an annual prepay and you’ve built a machine that quietly takes back what the customer paid for — and they do, eventually, notice. Roll the credits over, or true them up at renewal. Forfeiture is free margin today and a trust debit that comes due exactly when you’re asking them to sign again.

Raising it without a mutiny#

Sooner or later you will raise the effective price. Whether that lands as ordinary commerce or as betrayal turns on one thing the buyer can barely articulate but feels precisely: the reference price. People peg value to the number they were shown, and they acclimate to it fast — toothpaste or enterprise software, within a purchase or two the disclosed price simply becomes “the price.” Move against that anchor and you’re not adjusting a figure; you’re breaking a frame the customer had stopped questioning.

Here’s what most people miss: your sales motion decides how gently you can reprice, because it decides which anchor the customer holds.

If you sell the negotiated way, the whole “contact us” motion rests on a shared fiction both sides accept — there is a list price, and we will get to a real number that works. Discounts of 20–30% off that list are routine; 50% and beyond is not rare. That gap is a gift, because it hands you the kindest lever in all of pricing: hold the list, and let the discount decay across renewals as the account matures and wires itself into a dozen workflows. The reference — the list — never moves, so “your discount is a little smaller this year” lands far softer than “we raised the price,” and a locked-in customer will not rip out a system they depend on over ten or twenty points of discount. HubSpot does exactly this, even down-market: its own pricing announcement names “expiring discounts” as one of the things that will raise a customer’s bill at renewal. The list anchor holds, the effective price climbs, and nobody mutinies.

The disclosed meter has none of that cover. When the price is published and identical for everyone, that number is the promise, and there is no quiet discount to retire. To raise the effective price you have exactly two doors. You can move the visible number — and own it, in public. Or you can move the burn behind a frozen sticker and hope no one does the math. The second door is burn-rate repricing, and it is the disclosed world’s standing temptation precisely because it has no gentler lever to reach for. It is also its deepest trap: in a disclosed-price world the buyer staked everything on that one number, so the day they discover it quietly stopped meaning what it used to, the breach feels total. The naked meter can be raised honestly or covertly. There is no third, painless option — and the covert one only defers the bill, with interest.

So decide which world you’re in before you need the money. The negotiated wrapper buys you years of gentle, list-anchored increases. The disclosed meter commits you to repricing in the open — which is survivable, even routine, done with a little craft, and fatal the moment you reach for the shortcut.

And the craft is not complicated, in either world. When you move a price the customer can see: grandfather the people already paying you — their reference price is the most expensive one to break, and the cheapest to keep. Give real notice — a surprise is an insult; a heads-up is a courtesy, and the same number lands completely differently depending on which one it arrives as. State the why — “our costs rose” is a reason a buyer can actually accept now that everyone knows inference isn’t free; the COGS story is your alibi, so use it. And the one rule under all the others: never let the bill move in silence. Cursor’s economics weren’t the scandal; the silence was. The model survived the day they finally explained it.

The free tier is a CAC line, not a gift#

One meter setting gets mis-filed as generosity, and the mistake is expensive: the free tier. It is not generosity. It is an acquisition cost you pay in inference on people who have given you nothing yet — every free generation is real money out the door, and “free cadence,” a daily grant versus a monthly one, is an acquisition dial wired straight to your COGS.

Filed as a gift, the free tier does two bad things at once: it bleeds you quietly, and — worse — it satisfies the user so thoroughly they never need to pay. Filed as CAC, it has a job and a budget. Its job is to make the value undeniable and then create the need to upgrade; its budget is sized like the marketing spend it actually is. Set the ceiling to bite right where the value becomes obvious, not comfortably past it, and measure the thing that matters — conversion — not the applause from people who will never pay. What that subsidy costs you a month is precisely the kind of number worth modeling on your own inputs; that’s what the pricing calculator is for.

The manner is the whole game#

A price you set is an event. A meter you run is a relationship — and relationships are lost in the manner, not the math. Set the four levers so the meter feels fair on a day when nobody’s thinking about it. Choose your motion knowing which repricing future it hands you — the gentle list-anchored climb, or the open raise you’ll have to do with care. And when the day comes to ask for more, and it will, raise it in the light: grandfather the faithful, name the reason, and never move the bill while pretending you didn’t.

Change the number in the open, and customers grumble and renew. Change it in the dark, and they don’t grumble at all — they just leave, the moment they find out. That, in the end, is the whole difference between a price increase and a mutiny.