Call center KPI guides
Eighteen metrics, each with the formula, the current benchmark, and — the part most KPI pages leave out — the metric it will quietly damage if you optimise it on its own.
In short
A call center KPI is a quantified measure of contact center performance. The core set splits into four groups: speed (service level, ASA, abandonment), efficiency (AHT, occupancy, shrinkage, cost per contact), quality (FCR, CSAT, CES, transfer rate) and outbound (connect rate, RPC, list penetration).
Almost every one of them can be improved in isolation by damaging another. That is why each entry below names its paired metric.
How many KPIs to track
Fewer than you think. A contact center that reports 30 metrics weekly is usually managing none of them, because nobody can hold 30 numbers in mind well enough to notice when two of them move in opposite directions.
A workable core is six: one speed metric, one efficiency metric, one quality metric, one cost metric, one workforce metric and one outbound metric if you dial. Everything else is diagnostic — you pull it when the core six tell you something is wrong, not on a schedule.
A reasonable default core
Service level, AHT, FCR, cost per contact, occupancy, and connect rate for outbound. Six numbers, reviewed weekly, each with a named owner.
The 18 metrics
Benchmarks below are 2026 industry aggregates. They vary by vertical — see the by-industry tables before comparing yourself to any of them.
Service level (SL)
The percentage of contacts answered inside a target time, written as answered/seconds.
- Formula
-
(contacts answered within threshold / contacts offered) × 100 - Benchmark
- 80/20 is the classic default. 2026 voice queues commonly sit between 75/30 and 85/20.
- Read it next to
- Abandonment rate. Service level is a cost decision, not a standard — a higher target buys headcount.
Average speed of answer (ASA)
The average time an inbound caller waits in queue before an agent picks up.
- Formula
total queue wait time / contacts answered- Benchmark
- 25–28 seconds globally; under 15 seconds is top quartile.
- Read it next to
- Abandonment. A 25-second average can conceal a long tail where a third of callers give up.
Abandonment rate
The share of connected contacts that end before an agent handles them.
- Formula
(abandoned contacts / contacts offered) × 100- Benchmark
- 2–5% is healthy. Sustained above 8% needs intervention.
- Read it next to
- Average speed of answer. On outbound, abandonment is also a compliance ceiling, not just a service metric.
Average handle time (AHT)
The average total time an agent spends on one contact, start to finish.
- Formula
(talk time + hold time + after-call work) / contacts handled- Benchmark
- 4–7 minutes on voice. Complex B2B support runs 10–12; collections closer to 4.
- Read it next to
- First call resolution. This pairing matters more than any other in the list.
After-call work (ACW)
Time spent finishing a contact once the customer has disconnected — notes, dispositions, CRM updates.
- Formula
total ACW time / contacts handled- Benchmark
- Highly process-dependent. Track the trend, not an absolute target.
- Read it next to
- Disposition quality. Cutting ACW by simplifying wrap-up codes corrupts your campaign reporting.
First call resolution (FCR)
The percentage of contacts fully resolved on the first interaction, with no follow-up needed.
- Formula
(contacts resolved first time / total contacts) × 100- Benchmark
- 70–74% industry average; 80%+ is top-tier. Retail ~78%, healthcare ~71%, financial services 67–71%.
- Read it next to
- Average handle time. Every point of FCR improvement removes roughly a point of repeat volume.
Transfer rate
The share of contacts an agent passes to another agent, team or support tier.
- Formula
(transferred contacts / contacts handled) × 100- Benchmark
- No universal target. Rising transfer rate is the signal, not the level.
- Read it next to
- Routing accuracy and training. It is almost never an individual agent problem first.
Occupancy rate
The share of an agent's logged-in time spent handling contacts rather than waiting.
- Formula
((talk + hold + ACW) / logged-in time) × 100- Benchmark
- 75–82% is the healthy band; support medians sit near 78%.
- Read it next to
- Attrition, on a roughly 90-day lag. Above 85% the turnover cost cancels the seat-cost saving.
Shrinkage
The share of paid agent time unavailable for contacts: breaks, training, meetings, coaching, absence, downtime.
- Formula
(unavailable paid hours / total paid hours) × 100- Benchmark
- 30–35% is normal. Understating it is the most common cause of a staffing model that fails in week one.
- Read it next to
- Service level attainment. A model built on 20% shrinkage will miss every target it forecasts.
Schedule adherence
How closely agents work the hours they were scheduled, including breaks at the assigned times.
- Formula
(time in adherence / total scheduled time) × 100- Benchmark
- 85–90% typical; top quartile reaches roughly 92%.
- Read it next to
- Intraday service level. Adherence failures show up as afternoon queue spikes, not as a daily average.
Customer satisfaction score (CSAT)
The share of surveyed customers who rate an interaction positively, usually the top two boxes of five.
- Formula
(positive responses / total responses) × 100- Benchmark
- 75–85% typical; 85%+ is the 2026 target.
- Read it next to
- Occupancy first, then AHT inflation, then FCR. CSAT is a trailing indicator that confirms what those three already showed.
Net promoter score (NPS)
A loyalty measure from a single likelihood-to-recommend question on a 0–10 scale.
- Formula
% promoters (9–10) − % detractors (0–6)- Benchmark
- Varies enormously by industry; only useful against your own trend.
- Read it next to
- Nothing at contact level. NPS measures the relationship, so using it to score individual agents is a category error.
Customer effort score (CES)
How much effort the customer had to spend to get their issue resolved, usually on a 1–7 scale.
- Formula
mean of responses, or % low-effort responses- Benchmark
- No cross-industry benchmark worth quoting.
- Read it next to
- Repeat contact rate. CES predicts churn and repeat volume better than satisfaction does.
Cost per contact (CPC)
The fully loaded cost of handling one contact — labor, technology, telecom, facilities, overhead.
- Formula
total operating cost / contacts handled- Benchmark
- Roughly $2 self-service against roughly $13 assisted.
- Read it next to
- Containment quality. Deflecting a contact that then comes back twice has raised your cost, not lowered it.
Agent turnover
The share of agents who leave over a period, usually annualised.
- Formula
(agents departed / average headcount) × 100- Benchmark
- 40–45% annual is common, some operations above 60%. Median tenure 13–15 months.
- Read it next to
- Occupancy and tenure. Replacement is generally costed at $10,000–$20,000 per agent.
Connect rate
The share of outbound dial attempts that reach a live person.
- Formula
(live connects / dial attempts) × 100- Benchmark
- Entirely list- and reputation-dependent; benchmark against your own campaigns only.
- Read it next to
- Caller ID reputation. A connect rate that collapses overnight is a labeling problem, not a pacing problem.
Right party contact (RPC)
A connect with the specific person the record identifies, rather than whoever answered.
- Formula
(right-party connects / live connects) × 100- Benchmark
- The metric that actually matters in collections and any regulated outreach.
- Read it next to
- Data quality. RPC is a list metric wearing a performance metric's clothes.
List penetration
The share of a lead list dialed at least once in a campaign cycle.
- Formula
(records attempted / total records) × 100- Benchmark
- Should approach 100% within the cycle you designed.
- Read it next to
- Dial volume. High volume with low penetration means the dialer is re-hitting the same reachable pool.
Five ways KPIs get gamed
Not usually deliberately. These are the failure modes that appear when a metric becomes a target without its pair.
- AHT down, FCR down. Agents close faster by resolving less. Volume reappears next week as repeat contacts, and the cost per resolution rises even though cost per contact fell.
- Occupancy up, attrition up. Pushing occupancy from 78% to 88% looks like free capacity for one quarter, then the turnover bill arrives.
- Transfers down, holds up. If transfer rate is scored, agents keep calls they cannot resolve and park the customer on hold instead.
- ACW down, data quality down. The fastest way to shorten wrap-up is to pick the first disposition in the list. Now your redial logic and suppression rules are working from fiction.
- Containment up, resolution flat. The AI-era version. Deflection counts as containment whether or not the customer's problem went away. See containment vs resolution.
Which KPIs belong to which role
| Role | Owns | Should not be scored on |
|---|---|---|
| Agent | Adherence, QA score, FCR, CSAT | Service level, abandonment — they cannot control staffing |
| Supervisor | Team FCR, QA calibration, coaching cadence, transfer rate | Cost per contact |
| Workforce management | Forecast accuracy, shrinkage, occupancy, schedule adherence | CSAT |
| Operations lead | Service level, cost per contact, attrition | Individual agent AHT |
| Campaign manager | Connect rate, RPC, list penetration, conversion | Inbound service level |
Deep-dive guides in progress
Each of these expands one metric into a full page: worked calculations, the instrumentation you need to measure it honestly, and the interventions that actually move it.
- First call resolution: measurement methods and why they disagreeIn progress
- Average handle time: a decomposition worksheetIn progress
- Service level: choosing a target from cost, not conventionIn progress
- Occupancy and shrinkage: the staffing arithmeticIn progress
- Cost per contact: what belongs in "fully loaded"In progress
- CSAT, NPS and CES: which instrument answers which questionIn progress
- Agent turnover: a replacement-cost model you can populateIn progress
- Outbound metrics: connect rate, RPC and penetration togetherIn progress
These are listed without links on purpose — we would rather show you what is coming than send you to a page that does not exist yet. Tell us which to prioritise.
Frequently asked questions
What are the most important call center KPIs?
Six cover most operations: service level (are you answering fast enough), average handle time (how long contacts take), first call resolution (are they actually resolved), cost per contact (what it costs), occupancy (is the workload sustainable) and, if you dial out, connect rate.
FCR is the single most predictive of overall effectiveness, because it correlates directly with both satisfaction and cost per resolution.
How do you calculate average handle time?
Add talk time, hold time and after-call work, then divide by contacts handled. Excluding ACW is the most common error and it understates the real figure by 15–25% in most operations.
What is the difference between occupancy and utilization?
Occupancy measures contact-handling time against logged-in time. Utilization measures it against paid time, so it includes breaks, training and meetings.
Occupancy answers "how hard is this shift"; utilization answers "how much of what we pay for reaches a customer". Occupancy targets 75–82%; utilization typically 75–85%.
How often should call center KPIs be reviewed?
Service level and abandonment intraday, because you can still act on them today. The rest weekly. Re-baseline against external benchmarks quarterly.
Monthly-only reporting means every problem is discovered a month after it started costing money.
Should agents be scored on average handle time?
Not on its own. AHT is the easiest metric in the contact center to improve destructively — an agent can shorten it by ending calls before the issue is resolved, which raises repeat volume and total cost.
If you score AHT, score it jointly with FCR and QA, and treat a fast agent with poor resolution as a performance problem rather than a top performer.
Every metric here, reported automatically
DialedIn's dashboards calculate these KPIs per campaign, queue and agent, so you are comparing against your own trend rather than a blended industry average.