AI Summary

A KPI, or key performance indicator, is a number a business tracks to measure how well something is working.

Contact center KPIs fall into four groups. Access KPIs show whether callers get through: service level, average speed of answer, abandon rate, and blocked calls. Resolution KPIs show whether problems get solved: first call resolution, repeat caller rate, transfer rate, and customer satisfaction.

Workforce KPIs show whether staff are used well: average handle time, occupancy, utilization, shrinkage, and schedule adherence.

Quality KPIs cover quality assurance (QA) scores and sentiment. The most important call center reports are interval volume, queue performance, abandoned calls, call reasons, repeat callers, and agent performance.

A new manager should baseline 90 days of data and confirm how each metric is calculated before changing anything.

The New Manager’s Problem: Too Many Numbers, Not Enough Answers

Taking over a contact center comes with a strange welcome gift: a dashboard with 40 metrics, half green, a quarter red, and nobody sure who set the targets.

Meanwhile, the CFO wants to know why payroll is up and the agents want to know why they’re getting written up for handle time.

The metrics aren’t the problem.

The lack of hierarchy is. This guide sorts the contact center KPIs (often called call center KPIs or call center performance metrics) that drive decisions from the ones that just decorate the wall, explains the call center reporting worth running, and walks through the staffing math that ties it all together.

It’s written for supervisors, contact center managers, and operations leaders running a real queue with multiple agents.

For a front desk or small office team that’s just trying to answer the phones better, the simple guide to call center metrics and staffing is the better starting point.

The First 90 Days: Baseline Before Changing Anything

The fastest way to make a contact center worse is to start fixing it in week one.

Change everything at once, and there’s no way to tell which change did what. Build a baseline first.

Pull 90 Days of History

Ninety days covers enough Mondays, month-ends, and holiday weeks to see the real pattern.

Pull it by 30-minute interval, by queue, and by agent. That baseline becomes the measuring stick for every change that follows.

Audit How the Numbers Are Calculated

This is the step most new managers skip, and it’s the one that bites hardest.

Two contact centers can report the same abandon rate and be measuring completely different things. Before trusting any number, confirm:

• The short abandon threshold. Are hang-ups in the first 5, 10, or 20 seconds excluded? A center that quietly moves the threshold from 5 to 20 seconds can lower its reported abandon rate without changing anything a customer experiences.
• What counts as a call. Are after-hours calls, IVR-only calls, and calls to voicemail included or excluded?
• Callbacks. Is a call that converted to a callback counted as abandoned, answered, or both?
• Overflow and transfers. Does a call that overflows from one queue to another get counted twice?
• Wrap codes. Are agents actually using disposition codes, or is “Other” the most popular reason for calling?

Listen to the Calls

Numbers show what happened. Calls show why. Listening to 20 or 30 recorded calls, and spending ten minutes with each agent, tells a new manager more than a week of dashboards.

Just make sure recording follows the consent rules in each state where callers sit (Techmode’s guide to call recording laws by state covers the details).

Contact Center KPIs: The Four Groups That Matter

A KPI (key performance indicator) is a measurable number tied to a goal, such as answering 80% of calls within 20 seconds.

Most contact center and call center KPIs fit into four groups, and the order matters: access first, then resolution, then workforce, then quality.

Access KPIs: Can Callers Get Through?

These are the first numbers to check, because nothing else matters if customers can’t reach a person.

Service level is the share of calls answered within a target time. The traditional standard is 80/20, or 80% of calls answered within 20 seconds.

It’s a reasonable default, but not a law. SQM Group’s research found only about 16% of call centers hit 80/20 consistently, and many set the bar higher than their customers actually need.

The right target depends on the industry, the call type, and what customers will tolerate.

Average speed of answer (ASA) is the average time callers wait before an agent picks up. Watch the 90th percentile, not just the average.

An ASA of 25 seconds can hide a tail of callers waiting three minutes.

Abandon rate is the share of callers who hung up before reaching an agent, usually with short abandons excluded.

A common formula subtracts calls that ended within the first five seconds from both the top and bottom of the calculation, which TechTarget walks through with a worked example.

Industry guidance commonly treats 5% to 8% as typical. Here’s the catch: service level and ASA only measure calls that were answered. A queue can post a great service level while losing a third of its callers, because the impatient ones aren’t in the math anymore.

Always read abandon rate alongside service level.

Blocked calls are calls that never reached the queue at all, because every trunk was busy or the IVR sent the caller to a dead end.

They rarely show up on a standard queue report, which is exactly why they’re dangerous.

Resolution KPIs: Are Problems Getting Solved?

First call resolution (FCR) is the share of issues resolved on the first contact, with no callback needed. SQM Group benchmarks the industry average at about 71%, with 70% to 79% considered good and 80% or higher world class.

FCR matters because every repeat call costs twice: SQM’s research found customer satisfaction drops about 16% for each additional call needed to resolve an issue.

The tricky part is measurement. Agent-reported FCR (“did you resolve it?”) is wildly optimistic. Survey-based FCR is better.

The most practical proxy is the repeat caller rate: the share of callers who call back about the same issue within 48 to 72 hours.

Transfer rate is the share of calls passed to another agent or queue. A high transfer rate usually points to a routing problem, not an agent problem.

Customer satisfaction (CSAT) and Net Promoter Score (NPS) capture how customers felt about the experience. Both depend heavily on survey design, so a CSAT that only surveys answered calls (and never the callers who gave up) will always look better than reality.

Sentiment is the newer addition. With AI transcription, every call can be scored for tone, not just the 2% that a QA team has time to review.

Workforce KPIs: Is the Team Used Well?

Average handle time (AHT) is talk time plus hold time plus after-call work. Always break it into those three parts. Long talk time points to call complexity or training.

Long hold time points to agents hunting for information. Long after-call work points to clunky systems or a CRM that doesn’t talk to the phone system.

AHT should be a planning number, never an agent target.

Agents graded on AHT learn to close calls fast without solving anything, and FCR quietly falls apart.

Occupancy is the share of logged-in time agents spend actually handling calls (talk, hold, and wrap) rather than waiting for the next one.

It sounds like a number to maximize. It isn’t. Call Centre Helper warns that sustained occupancy above 85% risks agent burnout, and small teams need even more slack because calls bunch up.

Utilization is the share of paid time agents spend logged in and available for calls.

Occupancy and utilization are often confused. Occupancy asks “how busy were agents while logged in?” Utilization asks “how much of the paid shift was spent logged in at all?”

Call center shrinkage is the paid time that never turns into phone time: breaks, meetings, training, coaching, vacation, sick days, and late arrivals. Call Centre Helper’s industry poll put average shrinkage at around 30%, and many centers run higher. Ignore it, and every schedule will come up short.

Schedule adherence is the share of time agents are doing what the schedule says they should be doing at that moment. Conformance is whether they worked the total hours scheduled.

An agent can conform perfectly and still have poor adherence by taking lunch during the rush.

Cost per call ties it to the budget, but only means something next to FCR. A cheap call that generates two repeat calls isn’t cheap.

Quality KPIs: Is the Work Good?

QA scores grade calls against a checklist. They work best when the checklist is short, calibrated across supervisors, and tied to coaching rather than discipline.

Compliance errors should be tracked separately from the overall QA score. A great call that skips identity verification isn’t a 90% call. It’s a problem.

KPI Formulas and Benchmarks at a Glance

KPI	Formula	Typical Benchmark
Service level	Calls answered within target time ÷ calls offered	80% in 20 seconds (traditional target)
Average speed of answer	Total wait time of answered calls ÷ calls answered	Set by target; watch the 90th percentile
Abandon rate	(Calls offered minus answered minus short abandons) ÷ (calls offered minus short abandons)	5% to 8% typical; 10% or more needs attention
First call resolution	Issues resolved on first contact ÷ total issues	About 71% average; 80% or more is world class
Repeat caller rate	Callers who call back within 48 to 72 hours ÷ unique callers	Lower is better; read with FCR
Average handle time	(Talk + hold + after-call work) ÷ calls handled	Compare within the same queue and call type
Occupancy	Handle time ÷ logged-in time	At or below about 85%
Utilization	Logged-in time ÷ paid time	Read alongside shrinkage
Shrinkage	Paid time unavailable for calls ÷ total paid time	Around 30%
Schedule adherence	Time in the scheduled state ÷ scheduled time	Set internally; track the trend

Call Center Reporting: The 9 Reports Worth Running

Call center reporting turns the raw call data a phone system collects into reports a manager can act on.

Each report below follows the same format: what it shows, what to look for, the red flags, and the action it drives.

1. Interval Volume Report

Shows: Calls offered by 15- or 30-minute interval, by day of week, over several weeks.

Look for: Peak intervals, the shape of each day, and differences between days of the week and weeks of the month.

Red flags: Peaks that don’t match the schedule, and volume patterns that changed while the schedule stayed frozen.

Action: This is the foundation of every staffing decision. Forecast from it, and build schedules interval by interval.

2. Queue Performance Report

Shows: Calls offered, answered, and abandoned per queue, with service level and ASA by interval. In TechmodeGO, this is Queue Performance Overview, Detailed Queue Statistics, and SLA Statistics.

Look for: Which queues miss target, and when.

Red flags: One queue consistently starved while another sits idle, and SLA breaches clustering in the same intervals.

Action: Rebalance skills, adjust queue overflow rules, or move agents between queues during known peaks.

3. Abandoned Call Report

Shows: Every abandoned call with the time, queue, and how long the caller waited before hanging up. TechmodeGO calls this Abandoned Queue Calls.

Look for: The wait-time curve. At what point do callers start giving up?

Red flags: Abandons spiking after a specific wait (often 60 to 90 seconds), and abandoned callers with no follow-up.

Action: Staff to stay under the abandon point, offer callbacks before it, and work the abandoned list.

4. Call Reason and Disposition Report

Shows: Why customers call, from wrap codes or AI-generated call summaries.

Look for: The top ten reasons, and which ones are preventable.

Red flags: “Other” as a top reason (meaning the codes don’t work), and a single preventable issue driving a big share of volume.

Action: This is the biggest lever for reducing volume. Fix the root cause upstream, improve self-service, or route routine reasons to an AI agent.

5. Repeat Caller Report

Shows: Callers who called back within 48 to 72 hours.

Look for: Repeat rate by call reason and by agent.

Red flags: One call reason with a high repeat rate, or one agent whose callers keep calling back.

Action: The honest check on FCR. Use it to find broken processes and coaching opportunities.

6. Agent Performance Report

Shows: Calls handled, AHT broken into talk, hold, and wrap, adherence, QA scores, and login history per agent. TechmodeGO’s versions are Agent in Queue Statistics and Agent Login History.

Look for: Trends over time and outliers against the team, not a ranked leaderboard.

Red flags: Very short AHT with high repeat calls, high hold time, and logged-in agents who rarely take calls.

Action: Targeted coaching. Read it next to the repeat caller report so speed is never rewarded at the expense of resolution.

7. IVR and Call Flow Report

Shows: Where callers go in the menu and where they drop off.

Look for: Options nobody picks, options everyone picks, and drop-off points.

Red flags: High hang-ups at a specific menu level, and callers zeroing out to reach a person.

Action: Simplify the menu, or replace it with an AI receptionist that lets callers say what they need.

8. Real-Time Wallboard

Shows: Calls waiting, longest wait, agents available, and today’s service level, live.

Look for: Early warning that the interval is going sideways.

Red flags: Supervisors managing from the wallboard all day instead of fixing the schedule that causes the spikes.

Action: Intraday moves: pull agents off projects, open overflow, or turn on callbacks. Historical reports drive the long-term fixes.

9. Trunk and Capacity Report

Shows: Peak concurrent calls against available channels.

Look for: How close peak usage gets to the ceiling.

Red flags: Hitting the channel limit, which causes busy signals that never appear on queue reports.

Action: Add capacity before the busy season, and confirm with the carrier that calls aren’t being blocked upstream.

Contact Center Staffing: Erlang C Without the Headache

Every contact center staffing model rests on one idea: calls arrive randomly, so a team needs more capacity than the raw workload suggests.

The tool most centers use to calculate that is Erlang C, named after the Danish engineer A.K. Erlang, who worked out the math for telephone exchanges more than a century ago.

Workload in Erlangs

One Erlang is one hour of call traffic. If a queue gets 390 calls in an hour and each call takes 4 minutes of handle time, that’s 1,560 minutes, or 26 hours of work arriving within one hour.

That queue has a workload of 26 Erlangs. In theory, 26 agents could handle it. In practice, 26 agents would mean callers waiting forever, because calls don’t arrive in a neat line.

What an Erlang C Calculator Shows

An Erlang C calculator takes call volume, handle time, and a service level target, then shows how many agents are needed to hit it.

The interesting part is how sharply results change with each extra agent. Call Centre Helper ran the numbers for a 26-Erlang workload:

Agents	Occupancy	Average Speed of Answer	Service Level (30 sec)
28	93%	55 seconds	56%
29	90%	28 seconds	72%
30	87%	16 seconds	82%

Going from 28 to 30 agents cuts ASA by more than two thirds and lifts service level by 26 points. Two agents. That’s why “just work harder” never fixes an understaffed queue, and why holding occupancy near 85% takes one more agent than the service level target alone would suggest.

The same math works in reverse. Every routine call an AI agent resolves comes straight out of the workload. If an AI agent handles 40 of those 390 calls an hour (hours, order status, appointment requests), that’s 160 minutes of work gone, and the queue drops from 26 Erlangs to under 24.

At the same service level, that’s roughly two to three fewer agents needed in that interval. For a look at what’s available, see this comparison of AI phone answering services.

From Agents on the Phone to Agents on Payroll

Erlang C answers how many agents need to be on the phones in a given interval. It doesn’t account for breaks, training, meetings, and sick days.

That’s where shrinkage comes in. The scheduling formula is simple:

Agents to schedule = agents needed ÷ (1 − shrinkage)

If Erlang C says 30 agents for the peak interval and shrinkage runs 30%, that’s 30 ÷ 0.70, or about 43 agents scheduled to have 30 on the phones.

Skip that step, and the center will be short in every interval, every day, and everyone will blame the agents.

Forecasting

Erlang C is only as good as the forecast feeding it. Start with the interval volume report, layer in seasonality and known events (billing cycles, holidays, weather), and compare forecast to actual every week.

Larger centers eventually move to workforce management (WFM) software. For many mid-sized centers, the phone system’s reports, a free Erlang C calculator, and a disciplined weekly review get most of the way there.

Metrics That Mislead

Every metric can be gamed or misread. These are the usual suspects:

• AHT as an agent target. Rewards speed over resolution and drives repeat calls.
• Service level without abandon rate. Looks great while customers leave.
• Monthly averages. Hide the Monday morning disaster inside a respectable monthly number.
• Agent-reported FCR. Almost always higher than what customers report.
• CSAT surveys sent only to answered calls. The angriest callers never get the survey.
• Leaderboards. Agents with the hardest call types always look worst.

A Review Cadence That Actually Gets Used

Cadence	Review	Owner
Real time	Wallboard: calls waiting, longest wait, agents available	Supervisor on duty
Daily	Yesterday’s service level, abandons by interval, abandoned callers to follow up	Supervisors
Weekly	Forecast versus actual, adherence, repeat callers, top call reasons	Contact center manager
Monthly	KPI trends, staffing model, shrinkage, QA calibration, coaching plans	Manager and leadership
Quarterly	Targets, routing design, IVR or AI receptionist changes, capacity	Leadership with the provider

How Techmode Supports Contact Centers

A contact center lives or dies on how it’s configured: queues, skills, overflow, callbacks, and the reports that show whether any of it works.

Techmode has years of experience with call centers and contact centers, so it brings proven best practices to every setup instead of handing over a blank admin panel.

Every deployment starts with Premier Launch, Techmode’s white-glove installation, where a dedicated project manager and install team map current call patterns, build queues and skills around the real interval peaks, configure SLA targets and short abandon thresholds so the reports mean what they say, and test every call flow before go-live.

After launch, Techmode’s Concierge Services team stays involved. Concierge Service is always around to help evaluate and refine the contact center, free of charge: reviewing queue and agent reports with the manager, spotting staffing gaps, and adjusting queues, overflow, skills, and callbacks as volume shifts.

Concierge Service is U.S.-based and available 24/7, with no offshore call center, and every change is covered by the lifetime configuration guarantee.

TechmodeGO’s contact center includes the reports covered in this guide out of the box: SLA Statistics, Queue Performance Overview, Abandoned Queue Calls, Queue Callbacks, Queue Answered Calls By Waiting Time, Agent in Queue Statistics, Agent Login History, and a live wallboard.

The AI phone system adds transcription, summaries, and sentiment on every call, which turns the call reason report from a wrap-code guessing game into something reliable.

Businesses choose the AI engine, including ChatGPT, Grok, a Techmode private AI server, or their own private server for centers that need call data kept in-house (the AI phone system page covers the privacy options).

Through the 3CX MCP server connection, managers can ask the phone system questions in plain English, such as “Which queue had the most SLA breaches last week, and which agents were logged in?”, instead of building a custom report.

And when routine calls are eating capacity, the AI Receptionist and AI agents are included in TechmodeGO (for a look at how real callers interact with one, see this AI agent case study).

Because Techmode is a CLEC and carrier of record, it owns the platform and the lines. That gives it carrier-level visibility into trunk capacity and blocked calls, the problems that never show up on a queue report.

Everything runs on private, triple-redundant AWS instances with a 99.999% uptime SLA, backed by an NPS of 85.7 across more than 1,000 post-support surveys (against an industry benchmark around 31) and an A+ BBB rating.

Taking over a contact center or trying to fix one? Talk to the Techmode team about a free review of current queue performance and staffing.

Frequently Asked Questions

Q: What are the most important contact center KPIs?

The most important contact center KPIs are service level, average speed of answer, abandon rate, first call resolution, repeat caller rate, average handle time broken into talk, hold, and wrap, occupancy, shrinkage, schedule adherence, and customer satisfaction. Together they cover whether callers get through, whether problems get solved, and whether staff are used well. Most centers should track these first before adding anything else.

Q: What is a good service level for a call center?

The traditional benchmark is 80/20, meaning 80% of calls answered within 20 seconds. It’s a reasonable default, but SQM Group found only about 16% of call centers hit it consistently, and the right target depends on industry, call type, and customer expectations. Service level should always be read alongside abandon rate, because it only measures calls that were answered.

Q: What is the difference between occupancy and utilization?

Occupancy is the share of logged-in time agents spend handling calls, including talk, hold, and after-call work. Utilization is the share of paid time agents spend logged in and available at all. Occupancy above about 85% for long stretches is widely treated as a burnout risk, while low utilization usually points to high shrinkage or loose scheduling.

Q: How does an Erlang C calculator work?

An Erlang C calculator takes call volume, average handle time, and a service level target for an interval, then estimates how many agents need to be on the phones to hit that target. It accounts for calls arriving randomly, which is why the answer is always higher than the raw workload. The result covers agents on the phones only, so schedules must add shrinkage on top.

Q: What is call center shrinkage and how much should be planned for?

Call center shrinkage is paid time that doesn’t turn into phone time, including breaks, meetings, training, coaching, vacation, sick days, and late arrivals. Industry surveys put it at around 30% of paid time, and many centers run higher. To schedule correctly, divide the agents needed by one minus the shrinkage rate, so 30 agents needed at 30% shrinkage means about 43 scheduled.