Before you buyIntermediate· 6 min read

What metered pricing really costs

The one thing to remember

Multiply the measured cost per run by your realistic monthly volume, then double it for the runs you will not plan.

The question

Convert a per-run price into the number that will actually appear on your card.

Figure

How What metered pricing really costs works, in one picture

The same argument as the text, as a chain. Each step is what makes the next one possible.

1
Find the measured cost, not the advertised one
A listing run in the sandbox carries a measured cost per run, from a real execution on a real task, which is why it is often higher than the figure in a maker's own documentation. No listing has been run there yet: the sandbox waits on isolated infrastructure.
Where the field is blank, we have not measured it, and a blank should make you more cautious about a metered listing rather than less.
2
Estimate volume from what you do now, not from what you plan
The reliable method is to count the thing the agent replaces. If it drafts replies to support tickets, your volume is your ticket count, not your optimism about automation.
Then add the runs nobody budgets for: retries, the ones that produce output you throw away, and the week somebody discovers it and runs it on everything.
Retries are the most commonly missed line. A task set with a 57 percent success rate runs nearly twice.
3
Compare against the subscription version of the same job
A subscription at a fixed monthly price and a metered listing at a fraction of a penny per run are the same purchase with the risk placed differently. Metered is cheaper when volume is low and predictable, and worse when it is neither.
Work out the crossover volume. If your realistic number is anywhere near it, take the fixed price and buy the certainty.

Try it

You have got it when

You have a monthly figure for a metered listing you were considering, including retries.

1
Find the measured cost, not the advertised one
A listing run in the sandbox carries a measured cost per run, from a real execution on a real task, which is why it is often higher than the figure in a maker's own documentation. No listing has been run there yet: the sandbox waits on isolated infrastructure.
Where the field is blank, we have not measured it, and a blank should make you more cautious about a metered listing rather than less.
2
Estimate volume from what you do now, not from what you plan
The reliable method is to count the thing the agent replaces. If it drafts replies to support tickets, your volume is your ticket count, not your optimism about automation.
Then add the runs nobody budgets for: retries, the ones that produce output you throw away, and the week somebody discovers it and runs it on everything.
Retries are the most commonly missed line. A task set with a 57 percent success rate runs nearly twice.
3
Compare against the subscription version of the same job
A subscription at a fixed monthly price and a metered listing at a fraction of a penny per run are the same purchase with the risk placed differently. Metered is cheaper when volume is low and predictable, and worse when it is neither.
Work out the crossover volume. If your realistic number is anywhere near it, take the fixed price and buy the certainty.

Try it

What a metered listing actually costsInteractive

Runs you actually need each month2000

Measured cost per attempt4p

Success rate on your work57%

Retries before you give up2

Fixed-price alternative (£/month)99

Advertised

£80

Actually billed

£129

Attempts per task

1.61

Surprise multiple

1.61x

Metered, billed on attempts Fixed price Crossover

The fixed price wins above 1,533 tasks a month. You are at 2,000, which is past it. The reason to care is that the crossover moves the moment your success rate does. Billed attempts this month: 3,230 at 4.0p each.

Drag the success rate down. The bill moves and the advertised price does not, because a metered vendor bills attempts and you were counting completions.

You have got it when

You have a monthly figure for a metered listing you were considering, including retries.

What metered pricing really costs

How What metered pricing really costs works, in one picture

Find the measured cost, not the advertised one

Estimate volume from what you do now, not from what you plan

Compare against the subscription version of the same job

Read next

What metered pricing really costs

How What metered pricing really costs works, in one picture

Find the measured cost, not the advertised one

Estimate volume from what you do now, not from what you plan

Compare against the subscription version of the same job

Read next