Cost per resolution, explained
Cost per resolution is the metric vendors use to make pricing sound rational: you pay for outcomes. But the number is only as honest as the definition of "resolution" behind it, and most definitions count deflection, tickets where the customer simply gave up. This page explains what cost per resolution really measures, why it misleads, and how to compute a version you can trust.
What the metric means
Cost per resolution is your total AI cost divided by the number of resolutions. As a pricing unit it sounds fair, because it ties spend to outcomes. The catch is definitional. If "resolution" means "the ticket did not reach a human," then you are paying for deflections, and a deflection where the customer gave up is not a solved ticket.
Gartner 2026 found only 9 percent of customers say a chatbot fully resolved their issue, so a resolution count based on containment can be wildly higher than the count of tickets actually solved.
Why it misleads
Two distortions inflate the apparent value and hide the real cost. First, deflection counted as resolution means you pay for non-solutions. Second, reopens are ignored: a ticket counted resolved that comes back within 72 hours was never resolved, but the meter already charged you. So the naive cost per resolution understates your true cost per solved ticket, sometimes badly.
One caution on any cost-per-resolution math: a "resolution" is only worth paying for if it actually solved the customer. Resolve reports verified resolution (confirmed by survey, confirmation, or no 72-hour recontact) separately from deflection, and treats reopen rate as the lie detector. A per-resolution bill that counts deflections, where the customer simply gave up, charges you for tickets that were never really solved. Grounding the price in verified resolution is the only honest way to compare cost.
A true cost per verified resolution
Compute it on verified resolution: a ticket confirmed by survey, confirmation, or no 72-hour recontact. Total AI cost divided by verified resolutions is your real unit cost. It will be higher than the marketed cost per resolution, because the denominator is smaller and honest. Under flat pricing this number falls as your verified resolution rate rises, which is exactly the direction you want, whereas a per-resolution meter keeps your true unit cost roughly fixed no matter how good the AI gets.
Frequently asked
What is cost per resolution?
Total AI cost divided by the number of resolutions. It sounds like outcome-based pricing, but it depends entirely on how "resolution" is defined. If it counts deflection, tickets where the customer gave up, the number flatters the vendor and understates your true cost per solved ticket.
Why is cost per resolution misleading?
Because it usually counts deflection as resolution and ignores reopens. Gartner found only 9 percent of customers say a bot fully resolved their issue, so a containment-based count can be far higher than the count of tickets actually solved. Compute it on verified resolution instead.
How do I get a true cost per resolution?
Divide total AI cost by verified resolutions, tickets confirmed by survey, confirmation, or no 72-hour recontact. It will be higher than the marketed figure but honest. Under flat pricing this true unit cost falls as your resolution rate improves.