Cost per resolution
In one sentence
Cost per resolution is the total cost of actually solving a customer problem across every attempt and every channel it took, divided by resolved issues, and it is the honest version of cost per call because it counts the failed attempts, escalations and repeat contacts that cost per call conceals.
Not to be confused with Cost per call.
Definition
Cost per resolution is what it actually costs to sort a customer problem out, counting every attempt.
It is the number that tells you whether automation saved you money, because cheap attempts that do not fix anything are not savings.
Cost per resolution cannot be gamed by the tactics that flatter cost per call, which is precisely why it is less commonly reported. Where the two disagree, the resolution figure is the one describing reality.
How it is calculated
- Total cost across all interactions relating to an issue, divided by resolved issues.
- Includes failed attempts, escalations and repeat contacts.
- Requires linking interactions to issues, which is the measurement difficulty and the reason it is under-used.
Why it exposes what cost per call conceals
- An automated interaction that fails and escalates costs the automation plus the full human handling.
- A cheap interaction that generates a repeat contact costs twice.
- Deflection achieved by friction shows up here as unresolved volume returning through another channel.
- A system optimizing cost per call can raise cost per resolution, and frequently does.
The calculation in practice
- Requires issue-level tracking across channels, which many organizations lack.
- Where full tracking is unavailable, repeat contact rate within a window is a workable proxy.
The on-site version
- A website voice agent that answers a question prevents an issue from becoming a contact at all.
- The counterfactual is difficult to measure, which is the honest limitation.
- The defensible measurement is total inbound contact volume per thousand visitors, which should fall if on-site answering works.
Common misconception
That cost per resolution is a refinement of cost per call. It frequently produces a materially different conclusion, and where the two disagree the resolution figure is the one describing reality.
Why it matters commercially
This is the metric that determines whether an automation program actually saved money. Organizations that report only cost per call can be confident about savings they did not make.
In voice specifically
A voice agent that handles the simple portions of a complex issue and escalates the rest fragments a single issue into several contacts. Cost per call falls because each contact is smaller, while cost per resolution rises because the problem now takes more interactions, and each voice escalation is dearer per minute than a text one, so the fragmentation hurts more.
Where AsqVox fits
The honest on-site measurement is whether total inbound contacts fall while satisfaction holds, rather than how many conversations the widget handled. AsqVox logs each conversation with its transcript and sentiment, which is what lets you check whether an on-site answer prevented a contact or merely delayed one.
Visual
Cheap per attempt, expensive per outcome
A system optimizing cost per call can raise cost per resolution. It frequently does.
Statistics
Every figure carries its source and year. Vendor numbers are labelled as vendor numbers, and where no reliable figure exists this page says so rather than borrowing one.
A human-handled call costs roughly USD 7 to USD 12 against roughly USD 0.40 for an agent-handled call.
USD 7 to 12 vs USD 0.40industry rangeWidely cited industry range, 2025 - An industry range rather than an audited figure, varying substantially by geography and complexity. These are the component costs that accumulate across a multi-attempt resolution.
All-in production voice agent cost lands around USD 0.11 to USD 0.33 per minute.
USD 0.11 to 0.33 per minindustry rangeIndustry-reported production cost range, 2026 - The delivered per-minute cost. It is the input a per-attempt automation cost is built from before it accumulates across a resolution.
Containment rates run 70 to 80 percent for mature deployments, 40 to 55 percent on average, and under 35 percent for rule-based systems.
70 to 80% mature, under 35% rule-basedindustry rangeIndustry-reported ranges, 2026 - Directional industry bands. The unresolved remainder that escalates carries both the automation cost and the full human cost, which is exactly the volume cost per resolution counts and cost per call does not.
There is no reliable cross-industry cost per resolution benchmark, and the metric is reported far less often than cost per call precisely because it requires issue-level tracking.
-no reliable figureWhich is why it cannot be borrowed. The defensible figure is your own, and computing it at all is a maturity signal.
There is no published methodology standard for linking interactions to issues across channels.
-no reliable figureThe measurement difficulty that keeps the metric under-used. Where full tracking is unavailable, repeat contact rate within a window is a workable proxy.
Examples
In practice
An organization celebrating falling cost per contact finds total support cost flat. Issue-level analysis shows contacts per resolved issue rising, because the automation handles simple portions of complex issues and escalates the remainder, fragmenting single issues into multiple contacts. Cost per contact fell because contacts became smaller, not because problems became cheaper to solve.
The everyday version
Cost per resolution is what it actually costs to sort a customer problem out, counting every attempt. It is the number that tells you whether automation saved you money, because cheap attempts that do not fix anything are not savings.
Usage
Who says it
- Finance and operations leaders with mature measurement.
- It appears in serious automation business cases and rarely in vendor material, because it needs issue-level tracking that flatters no one.
Where it turns up
- In an RFP, next to containment, resolution rate, escalation rate and repeat contact rate.
- In post-deployment reviews, where it is the number that settles whether the program saved money.
Common misuse
- Reporting cost per call as though it answers the same question, when the two frequently disagree.
- Failing to link interactions to issues, which makes the metric uncomputable.
- Excluding the automation build and maintenance cost from the calculation.
Questions people ask
How is cost per resolution calculated?
Total cost across all interactions relating to an issue, divided by resolved issues, including failed attempts, escalations and repeat contacts. The hard part is linking interactions to issues across channels, which many organizations lack. Where full tracking is unavailable, repeat contact rate within a window is a workable proxy.
What is the difference between cost per call and cost per resolution?
Cost per call is the cost of one interaction. Cost per resolution is the total cost of solving a problem across every attempt and channel, including failures and escalations. A system optimizing cost per call can raise cost per resolution, and frequently does, because an automated attempt that fails and escalates costs the automation plus the full human handling.
Why is cost per resolution reported less often than cost per call?
Because it requires issue-level tracking across channels, which many organizations do not have, and because it cannot be gamed by the tactics that flatter cost per call. There is also no reliable cross-industry benchmark and no published methodology standard for linking interactions to issues, so the defensible figure is always your own.
How do you measure cost per resolution for a website voice agent?
The counterfactual, a contact that never happened because the question was answered on the page, is difficult to measure directly, which is the honest limitation. The defensible measurement is total inbound contact volume per thousand visitors, which should fall if on-site answering works, measured against a satisfaction guardrail rather than by counting conversations handled.
Last reviewed 4 August 2026. Written and reviewed by Dhruv Dholakia, founder of AsqVox.