Customer support metrics are useful when they lead to a specific operational decision. Define how each measure is collected, compare similar work and review the underlying conversations before changing a policy or staffing plan.
These five measures cover speed, resolution, effort and customer sentiment. Use them together: a fast reply is not enough if the customer still needs help.
1. First response time
First response time measures the delay between a customer's initial request and the first qualifying response. Decide whether you measure elapsed time or business hours, and whether automated acknowledgements qualify. A receipt confirmation should not be confused with a useful answer.
Break the result down by channel and staffed hours. Report a median and a high percentile alongside the average so long waits are visible. Choose targets based on the service you promise and the resources available; an unsupported universal target such as “every email within one hour” does not account for your operating model.
When delays grow, inspect arrival patterns, routing and staffing before asking everyone to reply faster. Check answer quality at the same time.
2. First contact resolution
First contact resolution measures the proportion of eligible issues resolved in the first interaction. Define the interaction and the window used to detect a repeat contact about the same problem.
FCR = first-interaction resolutions ÷ eligible issues measured × 100
Keep the issue cohort consistent and account for reopened cases. A ticket marked closed does not automatically prove resolution. Our FCR guide explains how to investigate repeat contacts and test improvements.
Use issue-level results to find missing permissions, unclear policies or gaps in agent knowledge. Do not encourage agents to close a case before the customer's need is addressed.
3. Resolution time
Resolution time tracks the interval from the initial request to the defined resolution event. Document how your reporting handles reopened cases, waiting for a customer and dependencies on a carrier or another team.
Separate active handling time from elapsed time where the tools permit it. That distinction helps you see whether the problem is the amount of work or the delay between steps.
Review long-running cases by reason. Improve the relevant knowledge, permissions or escalation route, and include the changed workflow in agent onboarding. When evaluating automation through Chad, measure completed outcomes and exceptions alongside speed.
4. Customer effort score
Customer effort score (CES) asks customers how easy or difficult an experience was. Qualtrics' CES guide describes surveying effort around a particular interaction.
Keep the question, scale, direction and aggregation method consistent. A higher number can mean greater ease or greater difficulty depending on the survey, so label the result clearly. Include response count and response rate when interpreting changes.
Use comments and journey reviews to find unnecessary work: repeated identity details, unclear forms, transfers without context or a missing route to a person. Test the revised process from the customer's starting point.
5. Net Promoter Score
Net Promoter Score (NPS) uses a recommendation question on a zero-to-ten scale. In Bain's measurement method, ratings of nine or ten are promoters, seven or eight are passives, and zero through six are detractors.
NPS = percentage of promoters − percentage of detractors
The denominator includes all valid responses, including passives. For example, 60 promoters, 25 passives and 15 detractors among 100 responses produce an NPS of 45.
NPS reflects a broader customer relationship than a single support reply. Read the reasons behind responses and avoid attributing the score exclusively to the support team. Keep sampling and survey timing consistent when comparing periods.
Turn the review into action
For each metric, assign an owner, document its definition and choose a review cadence. When a result changes, look at issue mix, volume, survey participation and staffing before drawing a conclusion.
Select one observable problem and a change that addresses it. Check both the intended result and possible side effects, such as faster responses accompanied by more reopened cases. If performance changes during staffing transitions, include agent workload and churn in the investigation.
