Customer service KPIs are the handful of numbers a team reports to show how quickly, how completely and how well it answers customers. The twelve below cover speed, coverage, quality and outcomes, and each comes with its formula and the place it is read from in a shared inbox or help desk. A small team reports five or six of them weekly and the rest monthly, and it treats the speed numbers as hygiene rather than as the goal.
Report first response time and resolution time as a median and a 90th percentile, and reply coverage as the share of new conversations answered. Add the open backlog with the age of its oldest item, and service level attainment as the share of targets met. Add CSAT per teammate and inbox, first contact resolution and reopen rate monthly, and NPS quarterly. For an account team, add commitment follow-through, which is the share of promises kept on time. Public benchmarks exist only for consumer email response time, so every other target is set from the team's own first three months.
The 12 customer service KPIs at a glance
| KPI | Formula | Collected from |
|---|---|---|
| 1. First response time | Time of first team reply minus time of the customer's first message, per conversation, reported as median and p90 | Message and reply timestamps in the inbox or help desk |
| 2. Response time | Time of each reply minus time of the customer message before it, reported as median and p90 | Every reply event |
| 3. Resolution time | Time of close minus time of the first message, per conversation | Close or archive events |
| 4. Reply coverage | Conversations that received a reply ÷ new conversations, newsletters excluded, × 100 | New conversation and reply events |
| 5. Open backlog | Count of conversations awaiting a reply now, with the age of the oldest | The live unanswered view |
| 6. Volume | Messages received, messages sent and new conversations, per channel and per teammate | Message counts by inbox and sender |
| 7. Service level attainment | Targets met ÷ targets measured × 100, per policy, teammate and tag | Service level results |
| 8. CSAT | Satisfied responses ÷ total responses × 100 | Survey responses after close |
| 9. First contact resolution | Conversations closed after one team reply ÷ conversations closed × 100 | Reply counts per closed conversation |
| 10. Reopen rate | Conversations reopened within 7 days of close ÷ conversations closed × 100 | Reopen events |
| 11. NPS | Percentage of promoters minus percentage of detractors, from a 0 to 10 survey | A scheduled survey, outside the inbox |
| 12. Commitment follow-through | Promises kept by their date ÷ promises made × 100 | Commitments extracted from replies, or a list kept by hand |
Speed metrics: response and resolution time
First response time is measured once per conversation, from the customer's first message to the team's first reply, and it is the number most policies are written around. Response time is measured for every reply, from the newest customer message before it, so it catches a conversation that started well and then stalled. Resolution time runs from the first message to the close. All three are reported as a median, which describes the typical case, and a 90th percentile, which marks the point the slowest tenth exceeded. An average hides the tail where customers are lost.
Two public figures give a sense of the expectation on consumer email, and both are set out with their sources in email response time benchmarks and policy. A Toister Performance Solutions survey found that a reply within one hour meets the expectation of 88 percent of consumers. SuperOffice's benchmark test found an average reply time of 12 hours and 10 minutes among companies that replied at all, and 62 percent never replied. Both measured consumer-facing support, so a team that answers named clients sets its own target instead. A usual one is a first reply within one working day on email and two working hours on WhatsApp.
Coverage and load: reply coverage, backlog and volume
Reply coverage is the share of new conversations in a period that received any reply, with newsletters and notifications left out of the count. It is the metric that catches the request no one owned, which the speed metrics miss because an unanswered conversation has no response time. A conversation that started last month and was answered this month belongs to last month's coverage, so the figure for a closed month settles a few weeks after it ends.
The open backlog is a live count rather than a period figure. It shows how many conversations are waiting on the team right now, how many of those have no owner, and how old the oldest is. Volume is the count of messages received and sent and of new conversations, per channel and per teammate, and it is the denominator for most other figures. A rise in volume with flat staffing explains a rise in response time before anyone looks for another cause. A volume figure per teammate is the evidence for changing a rotation.
Quality metrics: service level attainment, CSAT and first contact resolution
Service level attainment is the share of targets met, where a policy sets a first reply, next reply and resolution target counted in business hours or around the clock. The clock pauses while the team waits on the customer. The figure is read per policy, per teammate and per tag, so a miss can be traced to a kind of request or a person. A team without a formal policy can still compute it against the target written in its response time policy.
CSAT is the share of survey responses that rated a conversation as satisfied. It is asked once when the conversation closes and read per teammate and per inbox with the response count beside it. What is CSAT has the formula with a worked example and the limits. First contact resolution is the share of conversations closed after a single reply from the team. It rises when the first reply carries the full answer rather than a question back. It is read with resolution time, since a high first contact figure with a long resolution time means the single reply took days to write.
Outcome metrics: reopen rate, NPS and commitment follow-through
Reopen rate is the share of closed conversations the customer reopened within a set window, such as seven days, and it is the check on a team that closes quickly. A low resolution time with a high reopen rate means conversations are being closed before they are resolved. NPS asks how likely the customer is to recommend the company on a 0 to 10 scale, a few times a year and outside any support thread. CSAT vs NPS sets out how the two are read together.
Commitment follow-through is the share of promises made in replies that were kept by their date, such as a document sent by Friday or a call booked for next week. It is the metric that predicts whether an account stays, and the one a team cannot read from a plain mailbox, because the promises live inside email. A tool that extracts commitments from replies reports it directly, and a team without one keeps a list by hand for its top accounts. Customer success metrics beyond response time describes it with two companion measures, open question age and accounts past a quiet threshold.
How to collect customer service KPIs from a shared inbox or help desk
Every help desk and most shared inbox tools record the events the formulas need: a message arriving, a reply sent, an assignment, a close and a reopen. The speed and coverage figures come from the built-in reports, read for a date range and filtered by inbox or teammate. Service level attainment needs a policy configured in the tool, and CSAT needs the survey switched on. Gmail and Outlook record none of this in a report, so a team on a plain shared mailbox exports the mail or adopts a tool that reads it.
The definitions matter as much as the tool. Two teams can report different first response times from the same mail. One may count an out-of-office reply as a response and the other not, or one may measure in business hours. A team writes its definitions down once, including what is excluded, and keeps them when it changes tools. One published example of such definitions is how analytics are calculated, and a team on another tool can ask its vendor for the same.
Which customer service metrics mislead small teams
The average response time misleads more than any other number. A single conversation that waited four days moves the mean of a small team by hours while the median barely shifts. CSAT misleads below a floor of responses, since one rating moves a teammate's score by ten points when there are ten responses in the month. Tickets per agent misleads when the work differs by inbox, as a billing conversation takes three times the effort of a password reset and the count treats them as equal.
First contact resolution misleads when it is counted from closes. A team that closes conversations after one reply and sees them reopened scores well on it and badly with customers. Response time misleads an account team when it becomes the goal rather than a hygiene number, as customer success metrics beyond response time argues. A fast reply that misses the promise it contained is worse than a slower one that keeps it. A small team reports the median and the p90, applies a response floor to CSAT, reads first contact resolution with reopen rate, and keeps response time in its place.
Where Sektra fits
Sektra's analytics report the first ten of these metrics from the events it records in a shared inbox. They cover volume, reply coverage, and first response, response and resolution time as average, median and 90th percentile. They also show the live backlog with urgent items that have no owner, per-teammate figures and service level results by policy, teammate and tag. CSAT comes from the surveys it sends when a conversation closes. Commitment follow-through comes from the commitments the assistant extracts from replies, which Lens, the relationship panel beside each thread, shows per client. NPS stays with the tool that owns the relationship, and help desks such as Zendesk and Freshdesk report a similar set on their middle plans, each with its own definitions.
Which KPIs to report and how often
Weekly, report first response time and resolution time as median and p90, reply coverage, the open backlog with the age of its oldest item, and service level attainment. Monthly, add CSAT per teammate above a response floor, first contact resolution read with reopen rate, and volume per channel and teammate. Quarterly, add NPS and, for an account team, commitment follow-through per client. Set each target from the team's own first three months on one set of definitions. Change a definition only with a note in the report that says the series restarted.
Common questions
What are the most important customer service KPIs?
The core set is first response time and resolution time as a median and a 90th percentile, reply coverage, and the open backlog with its oldest item. Service level attainment and CSAT complete it, and first contact resolution, reopen rate, NPS and commitment follow-through are added monthly or quarterly.
What is a good first response time for customer service?
Public figures exist for consumer email. A Toister Performance Solutions survey found that a reply within one hour meets 88 percent of consumers' expectations. SuperOffice measured an average of 12 hours and 10 minutes among companies that replied. A team that answers named clients sets its own target, such as one working day on email and two working hours on WhatsApp. It then measures the median and the p90 against it.
How do you calculate customer service metrics?
Each metric is a formula over events the inbox or help desk records. First response time is the first reply time minus the first message time, and reply coverage is conversations answered divided by new conversations. CSAT is satisfied responses divided by all responses, and service level attainment is targets met divided by targets measured, each multiplied by 100 where it is a share.
What is the difference between customer service KPIs and metrics?
A metric is any number the team can measure, such as messages received. A KPI is a metric the team has chosen to report against a target because it reflects the service the team is responsible for. The twelve in this guide are metrics, and a team picks five or six of them as its KPIs.
