How minutes work
Every plan on this site is priced by the minute, and that word carries real weight. This page is the one place it gets defined: what the meter runs on, what it never touches, what a minute actually pays for — and, stated just as plainly, exactly where the published definition stops today.
An allowance, then a continuation. Never a cliff.
Two things about a minute are settled structure, and they hold whatever the final figures turn out to be. First: included minutes are a monthly allowance — a fixed bucket a plan carries every billing month, the way a phone plan's own minutes work, not a cap that resets weekly or daily because that would read as more generous. Second: once a month's calls run past that allowance, the minutes past it are metered at a single per-minute rate — the same rate whichever plan you're on — rather than triggering a cutoff, a lockout, or a renegotiation. Going over doesn't stop the phone from being answered. It adds a line to that month's bill.
The rate itself, and every allowance it sits behind, is at the press with the rest of this cluster's figures. The mechanism is what this page can define; the numbers show up live, before you pay, the moment you start self-serve signup — or get confirmed in conversation at /access.
The meter's subject is the call itself.
A call runs from ring to goodbye, and the minutes it takes are what the meter measures — however the call ends. The two edges of that measurement, though, have no published rule yet, and this rail marks them that way instead of inventing one.
One minute, and everything standing under it.
The honest answer to "why isn't the price just what the AI costs to run" is that the AI answering the phone is one line item on a much longer list — and most of that list has nothing to do with any single call. A minute is a share of everything that has to be running, continuously, underneath every call for the call to work at all. Categories only, deliberately: no figure below, because the figures are at the press and the shape is what's true regardless.
A minute doesn't care how the call ended.
A call that ends as an order, a question answered, a transfer, or a callback costs the same for the same number of minutes — calls covers what each of those four endings means; this page needs only the shared fact. And yes, that answers the fair question directly: a three-minute call that never orders anything uses the same three minutes as one that does, because the line is open and staffed by the same system either way. Treating one as free would mean pretending its minutes cost nothing to run — which isn't true of either.
THE METER RUNS ON
THE METER NEVER TOUCHES
Months are built from unremarkable calls.
Picture one ordinary call: a caller asks whether a topping is available tonight, gets a straight answer, and places a small order while they're on the line — two items, one modifier, paid at pickup. Greeting to goodbye, a call like that runs about four minutes, and nothing about it is unusual.
Both figures in the ledger are stated examples, not measurements — move them and the shape of the arithmetic holds. The point is what it shows: a real month accumulates from calls exactly this unremarkable, stacking up day after day — not from one dramatic call blowing past an allowance on its own. The estimate runs the same arithmetic on numbers you supply.
Two open questions this page refuses to answer plausibly.
A restaurant asking "does the clock start on the greeting or when I start talking, and does thirty seconds round up or down" is asking a completely reasonable question this page cannot answer with a specific rule — because no such rule has been published anywhere. That's a real gap, stated as one, not a detail this page forgot.
The second question is sharper: if a call fails because something on Dohos's own side breaks mid-call, is that minute billed like any other? A lot of people would guess the answer on instinct — of course you shouldn't pay for our mistake. The instinct is reasonable, and it is also not something this page will state as policy, because confirming it honestly requires a precise, auditable definition of what counts as a Dohos-caused failure versus an ordinary dropped call — and that definition doesn't exist published anywhere either. Writing the comforting version here anyway would describe a policy about your bill that isn't actually settled. When the minute definition is pressed, this is exactly the kind of term that gets settled with it.
A flat rate has to be priced for someone's worst month or it loses money on it — and pricing every account for its worst month makes the rate too high for the account having an ordinary one. An allowance that scales with real volume, plus overage that never cuts a call off, stays fair to a quiet month and a busy one alike.
A pure per-minute model with no base would make an ordinary month's bill less predictable, not more — nothing about it would be known in advance. The allowance is what makes a typical month plannable; the metered continuation is what keeps an unusual month billed rather than refused. The two pieces do different jobs.
The costs a minute pays for — the whole engraving above — don't shrink to zero because one restaurant's volume is low this month. A free tier would need every paying plan to quietly absorb that cost instead. Same cost, moved somewhere less visible — not fairness.
Run your numbers with us.
Rates are being set at the press. Sign up to see the live rate card and check out yourself, or request access and we'll walk your call volume, your plan shape, and your final number together.