KPI benchmarking center

CSAT = ratings of 4 or 5 (on a 1–5 scale) ÷ all ratings × 100

Example: 180 ratings, 153 of them 4 or 5: 153 ÷ 180 × 100 = 85%.

How to read it: Read it with the response rate and by channel, agent and contact reason: one bad week on one topic often explains a drop.

Watch out: Some teams count only 5s (top box), which gives a lower number for the same service. Low response rates skew toward very happy and very angry customers.

NPS = % promoters (9–10) − % detractors (0–6), from −100 to +100

Example: 200 answers: 90 promoters (45%), 40 detractors (20%): 45 − 20 = +25.

How to read it: Follow the trend and read the comments of detractors; the score alone does not say what to fix.

Watch out: Small samples swing a lot. NPS mixes product, price and support, so it is a weak measure of one support team.

CES = average answer to "How easy was it to solve your issue?" (often 1–7), or % who agree it was easy

Example: Answers 6, 7, 5, 7, 4 on a 1–7 scale: 29 ÷ 5 = 5.8.

How to read it: Low effort scores point at repeat contacts, transfers and long forms; compare by contact reason.

Watch out: Scales differ (1–5, 1–7, agree/disagree), so only compare scores collected the same way.

FRT = time of first agent reply − time the ticket was created (report the median)

Example: Five tickets answered after 0.5, 1, 1.5, 3 and 9 hours: median = 1.5 h (the average, 3 h, is pulled up by one ticket).

How to read it: Use the median and the share answered within your target; split by channel, since chat and email have very different targets.

Watch out: Business hours and calendar hours give very different numbers. Auto-replies must not count as a first response.

Resolution time = solved time − created time (median, ideally excluding time waiting on the customer)

Example: Median of 10 solved tickets: 22 hours.

How to read it: Look at the slowest 10% of tickets and why they wait: escalations, missing information, other teams.

Watch out: Reopened tickets and "pending customer" time can hide long waits; say which one your number includes.

Service level = contacts handled within the target time ÷ all contacts × 100

Example: 1,000 calls, 820 answered within 20 seconds: 82% (target "80/20" met).

How to read it: Look at it per hour of the day: a daily average can hide the hours when the target breaks.

Watch out: Decide whether abandoned contacts count, and say so; it changes the number.

FCR = issues solved on the first contact ÷ all issues × 100

Example: 400 issues, 284 solved without a second contact: 71%.

How to read it: Track it by contact reason: low-FCR reasons are where training, macros or product fixes pay off.

Watch out: How you detect a repeat contact (same customer within 7 days? same topic?) changes the result; keep one definition.

AHT = (talk or chat time + hold time + wrap-up time) ÷ contacts handled

Example: 6,300 minutes over 1,200 contacts: 5.25 minutes (5 min 15 s).

How to read it: Use it for staffing, not as an agent target on its own: pushing it down can hurt FCR and CSAT.

Watch out: Chat agents handling several conversations at once make chat AHT hard to compare with phone.

Occupancy = handling time ÷ available (logged-in, not on break) time × 100

Example: 6.2 hours handling out of 7.5 available hours: 83%.

How to read it: Too low means idle capacity; sustained very high occupancy leaves no recovery time between contacts and tends to tire agents.

Watch out: It is not the same as utilization, which divides by paid time instead of available time.

Productivity = tickets (or contacts) solved ÷ agent working days (or hours)

Example: 520 tickets solved in 20 working days: 26 a day.

How to read it: Compare agents doing the same queue and mix; pair it with QA and CSAT so speed does not replace quality.

Watch out: Different ticket types take very different effort; a raw count rewards easy queues.

Attrition = leavers in the period ÷ average head count × 100 (monthly, or × 12 for a rough annual rate)

Example: 2 leavers, average head count 25: 8% that month.

How to read it: Split voluntary and involuntary leavers and look at tenure: early leavers point at hiring and onboarding.

Watch out: Monthly and annual rates are often mixed up; say which one you report.

Customer churn = customers lost in the period ÷ customers at the start × 100

Example: Start with 400 customers, lose 10: 2.5% monthly churn.

How to read it: Read logo churn and revenue churn side by side: losing many small accounts and losing one large one call for different actions.

Watch out: New customers gained in the period must not be added to the base.

NRR = (start ARR − churn − contraction + expansion) ÷ start ARR × 100; GRR = the same without expansion (at most 100%)

Example: Start $100k, churn $6k, contraction $2k, expansion $10k: NRR 102%, GRR 92%.

How to read it: GRR shows how well you keep what you have; the gap to NRR shows how much growth comes from existing customers.

Watch out: Count only customers that existed at the start of the period.

Cost per ticket = total support cost in the period (people, tools, overhead) ÷ tickets handled

Example: $42,000 monthly cost, 3,000 tickets: $14 a ticket.

How to read it: Follow it per channel and contact reason; deflecting simple tickets lowers it, but the remaining tickets get harder.

Watch out: Teams include different costs (management, tools, office); write down what is in your number before comparing.