Per-extension vs per-minute pricing

What each model is actually charging for

They are not two prices for the same thing. They are charges for two different things, which is why comparing the headline figures tells you almost nothing.

Per-extension is a subscription to capability. Per-minute is metered consumption. A business that hires ten people who each make two calls a month is punished by the first and barely noticed by the second — and a support team on the phone all day is the reverse.

The crossover, which is one division

There is a single number that decides this, and it is easier to find than vendors make it look.

Divide the monthly per-extension price by the per-minute rate. The answer is how many minutes a person has to talk each month before the per-extension model becomes the cheaper one. Below that number, metered wins. Above it, the seat wins.

Ten hours a month is twenty-eight minutes a working day. That is a useful sanity check: a sales or support desk clears it easily, while a warehouse, a clinic reception or a field team often does not come close.

Finding your two numbers

Both inputs are already sitting in systems you have. Do not estimate them — estimates run high, and you will be quoted against whatever you say.

1. Pull last quarter's total talk minutes from your current provider's bill or your PBX's call records. Use three months, not one, so a quiet or frantic month does not skew it.
2. Divide by three for a monthly figure, then by the number of people who actually make and take calls. That gives minutes per person per month — the number you compare against the break-even.
3. Separate inbound from outbound while you are there. Many plans price them differently, and a mostly-inbound team can be paying for an outbound allowance it never touches.
4. Note your peak concurrent calls too. It does not affect this comparison, but it is the number that decides channel capacity, and you will be asked for it.

If the gap between your per-person minutes and the break-even is small — within about twenty percent either way — the pricing model is not what should decide your choice. Pick on features, support and whether you can keep your numbers.

Why a vendor picks the model they pick

Worth understanding, because it tells you where their incentives sit.

• Per-extension gives the vendor predictable revenue and gives you a predictable bill. Both sides like that. The vendor's risk is a customer who talks enormously; yours is paying full price for seats that barely call.
• Per-minute shifts the risk onto you. The vendor earns more when you talk more, and your bill moves with your business — which is fine when your revenue moves with it too, and painful when it does not.
• Bundled plans are per-extension with an allowance attached. They look like both and behave like per-extension until the allowance runs out, at which point the overage rate is rarely the competitive one.

None of these is dishonest. But a vendor who only sells one model will frame your situation as the one their model suits, and that framing is worth noticing.

Most real systems end up being both

The clean comparison breaks down in practice, because a phone system is not one product. The platform, the calling and anything AI-driven have different cost shapes and are increasingly priced separately.

• The platform — extensions, routing, menus, recording, reporting — is capability, and capability is naturally a seat charge.
• The calling is consumption, and consumption is naturally metered. It can also be removed from the equation entirely by bringing your own carrier.
• An AI agent is metered by nature, because it bills for conversation time rather than for a person who might be on a call.

That is why a plan can be per-extension for the system and per-minute for the AI without being inconsistent. It is two different things charged two appropriate ways. The question to ask is not "which model do you use" but "which parts am I paying a seat for, and which parts am I paying by the minute".

What the comparison leaves out

The model is not the whole bill. Four costs sit outside it and routinely swamp the difference:

• Number rental, charged per DID per month regardless of use, and easy to accumulate without noticing.
• Minimum commitments — a minimum number of seats, or a floor on monthly spend, which quietly converts a metered plan into a fixed one.
• Onboarding and porting, sometimes free and sometimes not, and worth confirming in writing before you commit.
• The cost of being stuck. If the numbers belong to the vendor, the price of leaving is a migration rather than a configuration change — which is the strongest argument for keeping the carrier relationship yours.

Where each one is the wrong fit

• Per-extension is wrong when most of your users are occasional — a shift roster, a seasonal team, people who are reachable but rarely calling. You are buying availability you do not consume.
• Per-minute is wrong when you need a budget you can defend. A bill that moves with traffic is hard to forecast, and a bad month costs twice: once in business and once in telephony.
• Bundled is wrong at both extremes. Light callers subsidise heavy ones, and heavy callers hit overage rates that are rarely competitive.

And one that is true of all three: a pricing model is not a reason to accept a system that does not do what you need. The difference between models is usually a fraction of what one badly-handled month of missed calls costs.

See it on OwnPBX

• OwnPBX pricing, with live rates
• What is a cloud PBX?
• Bringing your own carrier
• Finding your own numbers