Resolution rate
In one sentence
Resolution rate is the share of interactions where the customer issue was actually solved, which is a different question from whether the conversation ended without a human, and it is the metric containment is routinely mistaken for.
Not to be confused with Containment rate, First contact resolution, or Task completion rate.
Definition
Resolution rate is the share of interactions where the customer issue was actually solved. Containment only measures whether the conversation ended without a human.
A conversation can end without a human and solve nothing. That gap is the whole reason to measure resolution separately.
Resolution and containment answer different questions and are routinely conflated, which is how automation programs report a success they did not achieve. Pinning the two apart is the first move in reading either number honestly.
The distinction
- Containment asks: did this interaction end without a human.
- Resolution asks: was the customer problem solved.
- A conversation can be contained and unresolved, which is the abandonment and dead-end case, counted as a success.
- A conversation can be escalated and resolved, which is a success that containment reporting counts as a failure.
How resolution is actually measured
- Direct survey, which is accurate and gets a low response rate.
- Repeat contact absence within a window, which is the practical proxy most teams reach for.
- Outcome tracking, where the issue has an observable end state such as a booking made or an order placed. Most reliable where it is available.
- Model-assisted transcript review, where a model assesses whether the stated need was met. Scalable and imperfect, and it should be validated against human review.
Why it is reported less often than containment
- Harder to measure.
- Usually lower, which is commercially unattractive.
- It requires linking interactions to outcomes, which many organizations cannot do.
On a website the question is whether the visitor got the answer they came for. Observable proxies include whether they stopped asking, converted, or returned to the same page later. Transcript review with sentiment is the accessible method for most deployments.
Common misconception
That containment approximates resolution. The gap between them is precisely the set of interactions where the system failed quietly, which is the most important set to know about and the one containment reporting is structurally unable to show.
Why it matters commercially
Resolution rate is what the business actually cares about. Reporting it, even though it is lower than containment, produces a roadmap that containment reporting cannot: the gap between the two numbers is the list of things to fix.
In voice specifically
On a call there is often a follow-up contact that signals the first one failed. On a website the signals are quieter, but they exist: the visitor who stops asking, converts, or comes back to the same page later. The transcript with sentiment is what makes them readable.
Where AsqVox fits
Transcripts and sentiment allow assessment of whether conversations resolved something, which is the honest measure for a website voice agent rather than counting the conversations that simply ended.
Visual
The two numbers, and the gap between them
Containment counts conversations that ended. Resolution counts problems that ended. The gap between them is where the system fails quietly.
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.
Containment rates run roughly 70 to 80 percent for mature deployments, 40 to 55 percent for average deployments, and under 35 percent for rule-based systems.
70 to 80% mature, 40 to 55% averageindustry rangeIndustry-reported ranges, 2026 - Industry-reported rather than audited. Resolution rate is typically lower than containment for the same deployment, because containment includes conversations that ended without resolution. Quoted here as the reference the gap is measured against.
There is no reliable cross-industry resolution rate benchmark for automated channels, and definitions vary substantially.
-no reliable figureAny resolution figure quoted at you needs its definition attached before it means anything. Without one, two published numbers are almost certainly measuring different things.
There is no published validation study of model-assisted resolution assessment against human review, despite this being an increasingly common measurement approach.
-no reliable figureA model can score whether the stated need was met at scale, but the score is only as trustworthy as its agreement with human review, and that agreement is currently unmeasured in the literature. Validate locally before reporting a model-assessed number.
There is no benchmark for resolution in website voice contexts.
-no reliable figureThe metric was built for contact centers. On a website the honest substitute is your own observable proxies: whether the visitor stopped asking, converted, or returned to the same page later.
Examples
In practice
A deployment reports 74 percent containment. Adding repeat contact analysis within seventy-two hours shows a fifth of contained conversations generated a follow-up, giving an adjusted resolution rate substantially lower. The team reports the lower figure and uses the gap as the roadmap, with the top three repeat-contact reasons becoming the next three automation targets.
The everyday version
Resolution rate is whether the customer actually got sorted, as opposed to whether the conversation simply ended. Containment counts anything that did not reach a person as a win, including customers who gave up. Resolution counts only the ones you actually helped.
Usage
Who says it
- CX leaders with mature measurement, who care about the outcome rather than the channel.
- It appears in serious automation reporting and rarely in vendor marketing, because it is usually lower than containment.
Where it turns up
- Next to the containment definition, repeat contact rate, measurement methodology and CSAT in an RFP.
Common misuse
- Using containment as a resolution proxy, which hides the exact set of interactions where the system failed quietly.
- Failing to define what counts as resolved, which makes any resolution figure unfalsifiable.
- Reporting model-assessed resolution without validating it against human review.
Questions people ask
What is the difference between resolution rate and containment rate?
Containment measures whether the interaction ended without a human. Resolution measures whether the problem was solved. A conversation can be contained and unresolved, which is the abandonment case that containment counts as a success, and it can be escalated and resolved, which is a success containment counts as a failure. The gap between the two numbers is where the system fails quietly.
How do you measure resolution rate?
Four ways, in rough order of reliability: outcome tracking where the issue has an observable end state such as a booking or an order, repeat contact absence within a window as the practical proxy, direct survey which is accurate but low-response, and model-assisted transcript review which is scalable but should be validated against human review. Many organizations cannot link interactions to outcomes, which is why resolution is reported less often than containment.
Is resolution rate the same as first contact resolution?
No, though they are often confused because both ask whether the issue was solved. First contact resolution adds a constraint: it was solved on the first contact, with no follow-up. Resolution rate drops that constraint and asks only whether it was solved at all, across however many interactions it took.
Why is resolution rate reported less often than containment?
It is harder to measure, it usually comes out lower, and it requires linking interactions to outcomes that many organizations cannot connect. All three make it commercially less attractive to publish. All three are also reasons to report it, because the lower honest number is the one that produces a real roadmap.
Last reviewed 4 August 2026. Written and reviewed by Dhruv Dholakia, founder of AsqVox.