WFM Metrics for Support Teams: Schedule Adherence, Shrinkage and Occupancy (2026)

Schedule adherence, shrinkage and occupancy are the three workforce management metrics that decide whether a support roster survives contact with the day. Call Centre Helper says shrinkage normally comes out between 30 and 35%, considers adherence of 85 to 95% acceptable, and puts typical occupancy at 80 to 85%. This guide gives the formula, a worked calculation and a fix for each, plus average speed of answer (ASA), and shows how all four move when AI agents resolve part of your volume.

It is written for support teams of 20 to 60 agents on chat and phone. For forecasting and Erlang C, start with our contact center workforce management guide; this one covers the metrics you review weekly.

Key takeaways

• Adherence measures timing, conformance measures hours. An agent who starts 5 minutes late and stays 5 minutes late scores 100% conformance but about 92% adherence, in Call Centre Helper's example.

• Apply shrinkage by dividing, never by adding. 70 agents needed at 30% shrinkage means 70 / 0.7 = 100 scheduled, not 91.

• Occupancy above 85% is a warning sign, below 70% is often just a small queue. At the same 80/20 service level, our Erlang C runs give 50% occupancy at 20 calls an hour and 87% at 400.

• ASA hides the worst waits. Abandoned calls do not count toward it, and a 20 second daily average can include a 2 minute lunch hour.

• When AI agents take the easy contacts, human AHT rises and occupancy falls. Re-baseline every target that was set before automation.

Which tools track these WFM metrics?

Lorikeet Workforce Manager builds the roster from ticket and call volume by channel, day and hour, and its coverage reports show where each week falls short of forecast demand, by channel and hour. It is built for support teams of 20+ agents running live chat or phones. Live adherence tracking is not part of the launch; the launch post says intraday management is next. Tools that track adherence during the day today:

• Assembled lists real-time adherence monitoring on every WFM plan, with Slack or email alerts when agents are out of adherence (Assembled pricing).

• Calabrio (now part of Verint) compares agent adherence with live ACD status; Verint pitches it at inbound contact centres moving off manual scheduling (Calabrio WFM).

• Playvox (now part of NiCE) reports punctuality, adherence and occupancy and is built for digital channels (Playvox WFM datasheet).

• Zendesk WFM offers real-time activity tracking of what agents are working on, a fit if you already run Zendesk (Zendesk WFM).

For pricing and fit side by side, see our comparison of workforce management software for support teams.

Workforce management metrics at a glance

What is schedule adherence and how do you calculate it?

Schedule adherence is the share of scheduled time an agent spent doing what the schedule said, when it said: minutes in adherence divided by total scheduled minutes, times 100. Call Centre Helper counts sanctioned activities such as meetings and coaching as adherent when they happen during their scheduled slot.

Worked calculation

An agent is scheduled 9:00 to 17:00, 480 minutes including a lunch break from 12:30 to 13:00. They log in at 9:07 and come back from lunch at 13:12.

• Minutes out of adherence: 7 (late start) + 12 (lunch overrun) = 19.

• Schedule adherence: (480 minus 19) / 480 x 100 = 96.0%.

• Conformance: if they stay until 17:19 to make up the time, they worked 480 minutes against 480 scheduled, so conformance is 100%.

That gap is the point. Call Centre Helper's occupancy and utilisation guide uses the same logic: an agent rostered 8:00 to 9:00 who works 8:05 to 9:05 has 100% conformance but is present for only 55 of 60 scheduled minutes, about 92% adherence. Conformance tells you whether you got the hours; adherence tells you whether you got them in the interval the forecast needed them.

What is a realistic schedule adherence target?

There is no universal benchmark, and the published ranges differ. Call Centre Helper says 85 to 95% is generally acceptable and that anything above 95% is very unlikely unless volume and AHT are extremely consistent. TechTarget says most contact centers aim for about 80% and treat 100% as unrealistic. Peopleware sees immature centres or those with long handle times at around 85%, and mature, short-handle-time centres targeting up to 95% or even 97%.

Whatever target you pick has a staffing cost. Peopleware's example: at a 90% adherence goal, an interval that needs 10 agents should be staffed with 11, because you should expect 10% of people to be out of adherence at any moment.

What to do when adherence is off

• Look at when, not just how much. Late starts, long breaks and early finishes need different fixes. Call Centre Helper recommends combining adherence with conformance data to see which.

• Build login time into the shift. If systems take up to ten minutes to open, schedule that time rather than counting it against the agent.

• Show agents their score the next day and discuss it in one to ones, as Peopleware suggests; a score seen weeks later changes nothing.

How do you calculate call center shrinkage?

To calculate shrinkage, divide total shrinkage hours (external plus internal) by total hours available and multiply by 100. Call Centre Helper defines external shrinkage as time away from the centre (holiday, sickness, lateness) and internal shrinkage as paid time at work but off contacts (breaks, training, meetings, coaching), and notes shrinkage is normally expressed across a full 12 months.

Worked calculation

Take a 30-agent team over four weeks, each agent paid 38 hours a week: 30 x 38 x 4 = 4,560 hours available.

Shrinkage = 1,025 / 4,560 x 100 = 22.5%. That is below the usual 30 to 35%, as you would expect from one quiet month. Call Centre Helper notes shrinkage peaks in the summer school holidays and around Christmas, and across the day between 9:00 and 11:00 and between 14:00 and 15:00, when meetings cluster. Dimension Data's Global Benchmarking Report, as cited by Call Centre Helper, gives an average of 35%.

How to apply shrinkage to staffing

Divide the interval requirement by (1 minus shrinkage). If Erlang C says an interval needs 22 agents and your shrinkage is 22.5%, schedule 22 / 0.775 = 28.4, so 29 people. Adding 22.5% instead (22 x 1.225 = 27) leaves you two short. Call Centre Helper's own example makes the same point at larger scale: 70 agents at 30% shrinkage need 100 scheduled, not 91.

What to do when shrinkage is off

• Measure it by interval and month, not as one annual number. A 30% average hides a Monday morning where meetings push it far higher.

• Move internal shrinkage into quiet intervals. Training, coaching and team meetings are the only shrinkage you fully control; schedule them where your occupancy is lowest.

• Do not cut training to hit the number. Call Centre Helper calls that probably a false economy.

What is call center occupancy, and how is it different from utilization?

Call center occupancy is the share of logged-in time agents spend handling contacts (talk, hold and wrap), while utilization is the share of shift time they are logged in at all. Call Centre Helper defines occupancy as total contact handling time divided by total logged time, or equivalently traffic intensity in Erlangs divided by agents, and utilization as total logged-in time divided by total shift time, which brings internal shrinkage like training into view.

Worked calculation: occupancy vs utilization

An agent's shift is 450 minutes once paid breaks are excluded. They spend 60 minutes in training and a team meeting, so they are logged in for 390 minutes, and they handle contacts for 320 of those.

• Utilization: 390 / 450 x 100 = 86.7%.

• Occupancy: 320 / 390 x 100 = 82.1%.

What is a good call center occupancy rate?

Call Centre Helper says occupancy typically sits between 80 and 85% and that consistently higher than 85% risks advisor burnout, while also stressing there is no best-practice benchmark. Peopleware advises not exceeding 90% and cites Call Centre Helper's estimate of an 83% industry average. Call Centre Helper's Erlang C worked example uses 85% as the maximum and warns AHT may rise above it.

Team size matters more than most targets admit. We ran Erlang C at a 6 minute AHT and an 80% in 20 seconds target:

A small phone queue that hits its service level will show low occupancy. That is not idle staff; it is the cost of answering fast with few people.

What to do when occupancy is off

• Above 85% for weeks: add hours in the intervals where it peaks, and check whether AHT and sick leave are already climbing.

• Low but service level on target: fill the gaps with email backlog, training or coaching, rather than cutting staff below what the queue needs.

• Never use it as an agent target. Call Centre Helper is clear that agents cannot control when contacts arrive, so occupancy is a planning metric only; agent productivity should lean on adherence.

What is average speed of answer, and what is a good ASA?

Average speed of answer is total customer wait time for answered calls divided by the number of calls answered, measured from when the call enters the queue after any IVR. Call Centre Helper puts the global average at 28 seconds and notes the 80/20 rule (80% answered within 20 seconds) is often cited alongside it. Their page prints the formula with a "x 100", which would turn seconds into a percentage; leave that out. For chat, the equivalent is your first response time.

Worked calculation

In a week, 1,200 calls are answered after 30,000 seconds of total queue time. ASA = 30,000 / 1,200 = 25 seconds.

Erlang C predicts ASA too: probability of waiting x AHT / (agents minus Erlangs). For 96 calls an hour at a 5 minute AHT (8 Erlangs), our runs give an ASA of 61 seconds with 10 agents, 25 seconds with 11 and 11 seconds with 12. One person is the difference between a good hour and a bad one, which is why adherence at peak matters so much.

What to do when ASA is off

• Report it hourly. Call Centre Helper describes a centre on target for the day while customers waited 2 minutes every lunch hour.

• Read it with abandon rate. Abandoned calls do not count toward ASA, so the worst waits drop out. Call Centre Helper's 2024 survey found 38% of contact centres say most customers abandon after one minute in the queue.

• Fix the interval, not the average. Rising ASA at the same times each week is a roster problem; check adherence and shrinkage in those intervals first.

How do WFM metrics shift when AI agents handle part of the volume?

When AI agents resolve the simpler contacts, the human queue gets smaller and harder, so AHT rises, occupancy falls at the same service level, and older targets stop making sense. Forrester senior analyst Christina McAllister described the AHT effect to No Jitter: if a bot contains the easy, quick contacts, average handle time goes up, and teams measured on both containment and AHT end up penalised for succeeding.

Metric by metric:

• AHT and ASA: expect handle time on the human queue to rise. Set new AHT baselines per contact type, and recheck ASA because fewer agents on a longer AHT behave differently in Erlang C.

• Occupancy: the table above shows the effect. If AI agents cut a 200 call hour to 100, occupancy at the same service level falls from 80% to about 71% before AHT changes. That is expected, not a sign of overstaffing.

• Schedule adherence: matters more. In a smaller queue each absent agent is a larger share of capacity.

• Shrinkage: freed-up hours are the natural place for training on the harder work now reaching people.

• Escalations: forecast the contacts AI agents hand over as their own stream. Lorikeet's AI agents resolve customer problems end to end across phone, SMS, chat, email and WhatsApp, and loop in your team when they cannot solve an issue; those handovers land in the human queue.

Our hub guide works through a full before-and-after staffing example in the section on how AI agents change the staffing maths, and our piece on scaling support without hiring covers the planning side. Workforce Manager plans the human side from the same volume data, with capacity planning that shows shortfalls up to 12 months ahead, factoring in growth.

What still needs a human?

• Setting the targets. 80/20, 85% occupancy and 90% adherence are conventions. The right numbers depend on what a wait costs your customers and how your agents are holding up.

• Reading adherence fairly. A score does not know an agent was on a vulnerable customer's call; a team lead does.

• Deciding what counts as shrinkage. Whether to include rare leave types or back-office work changes the number; Call Centre Helper warns about double counting if you also hire to cover them.

• Agreeing definitions. Utilization means different things in different tools. Write down your formula before comparing with contact center benchmarks.

A weekly WFM metrics review

1. Check ASA and abandon rate by hour, by channel.

2. For the worst intervals, check adherence and actual versus planned shrinkage.

3. Check occupancy by interval: above 85% means add hours, low with service level met means use the time well.

4. Recalculate shrinkage monthly and divide, do not add, when you apply it.

For the vocabulary, see our WFM glossary entry. If you want the roster built for you, Workforce Manager checks each draft week against contracted hours, rest between shifts, channel skills, approved leave and demand, then flags under-covered shifts and suggests qualified, available agents. You can import your roster model from a spreadsheet or helpdesk export and start a 30-day free trial.