WFM guideForecasting

Contact centre forecast granularity

Q: Should a contact centre forecast in 15-minute or 30-minute intervals?

Most voice contact centres use 30-minute intervals as the default, and it is the right choice for the majority. 15-minute intervals are worth using when: volume is high enough that each 15-minute bucket still carries a statistically meaningful number of contacts (roughly 30+ contacts per interval as a rule of thumb); demand is volatile within the hour (sharp peaks that a 30-minute average would smooth away); and the operation can actually act on 15-minute information (breaks and schedules built to 15-minute precision, real-time management at 15-minute resolution). 15-minute intervals are a poor choice when volume is low — at, say, 5 contacts per 15-minute interval, random arrival variation dominates and the forecast becomes noise; the Erlang C staffing requirement also becomes unstable and tends to over-staff because the small-numbers effect inflates the agents-per-interval requirement. 60-minute intervals are generally too coarse for voice because they average away the within-hour peak that drives the actual staffing need — but they are appropriate for email, back-office, and low-volume queues where work is deferred and a within-hour peak does not create a service failure. The principle: match the interval length to the volume per interval and to the speed at which the operation responds.

Interval length is a quiet but consequential decision. Too coarse and you average away the within-hour peaks that drive understaffing; too fine and each interval carries so little volume that the forecast becomes noise and the staffing maths goes unstable. The right granularity balances accuracy against stability — and it differs by channel.

The core trade-off: accuracy vs. stability

Finer intervals (15 min) → more accuracy, less stability

Capture within-hour peaks the average would hide, enabling precise staffing. But each interval pools less volume, so random arrival variation is a larger fraction of the total — the forecast is noisier and the Erlang C requirement less stable. At low volume, the small-numbers effect inflates the per-interval requirement and over-staffs.

Coarser intervals (60 min) → more stability, less accuracy

Pool more volume per interval, so the forecast is more stable and less noisy. But the average smooths away within-hour peaks — a 60-minute interval can look adequately staffed while the first 20 minutes breached SL badly. For a live queue, the hidden peak is exactly the thing that causes the service failure.

Rule of thumb: choose the finest interval at which each bucket still carries roughly 30+ contacts. Below that, the noise and small-numbers instability outweigh the accuracy gain.

Interval lengths and where each fits

15 minutes

Best for

High-volume voice and chat queues with volatile within-hour demand, where the operation manages and schedules at 15-minute resolution.

Strength

Captures sharp within-hour peaks that coarser intervals average away. Enables precise break placement and tight intraday management. Most accurate staffing where volume supports it.

Weakness

Each interval carries less volume → noisier forecast, less stable Erlang C requirement. At low volume the small-numbers effect inflates the requirement and over-staffs. More forecasting and scheduling effort.

30 minutes

Best for

The default for most voice contact centres. A good balance for medium-to-high volume operations.

Strength

Enough volume per interval for a stable forecast and staffing requirement, while still capturing most meaningful within-hour demand shape. Industry-standard — most WFM tools and ACD reports default to it.

Weakness

Can still average away very sharp sub-30-minute spikes (e.g. a peak triggered by a TV advert or an outbound campaign send). For most operations this is an acceptable trade-off.

60 minutes

Best for

Lower-volume queues, email/ticket and back-office work, and long-range planning where within-hour shape is not the point.

Strength

Maximum forecast stability — each interval pools the most volume, so the forecast is least noisy. Less forecasting and scheduling overhead. Fine for deferred work where there is no within-hour service failure.

Weakness

Averages away the within-hour peak for live queues — a 60-minute average can look adequately staffed while the first 20 minutes were severely understaffed and breached SL. Too coarse for real-time voice/chat at any meaningful volume.

Daily (1440 min)

Best for

Back-office case work and email where the SLA is measured in days, not minutes, and work is processed against a backlog rather than a live queue.

Strength

Matches the planning unit to the work: deferred case work is planned at daily/weekly capacity-vs-inflow resolution, not interval-by-interval. No spurious within-day precision where it has no operational meaning.

Weakness

Useless for any real-time queue. Only appropriate where work genuinely has no within-day service constraint.

Practical guidance

Match interval length to volume per interval

Take the lowest-volume operating interval you care about staffing accurately. If it carries 30+ contacts at 15 minutes, 15-minute is viable. If a 15-minute interval has only a handful of contacts, step up to 30 or 60 minutes for that queue.

Match interval length to how fast you can act

There is no point forecasting at 15-minute resolution if breaks, schedules, and intraday management all operate at 30-minute resolution. The forecast granularity should not be finer than the granularity at which the operation can actually respond.

You can forecast and staff at different granularities by channel

A blended operation might run voice at 30 minutes, chat at 15 minutes (more volatile), email at 60 minutes, and back-office case work daily. The interval length is a per-workstream choice, not a single global setting.

Coarser for long-range, finer for short-range

Long-range capacity planning (months ahead) does not need 15-minute granularity — daily or weekly totals suffice. Reserve fine granularity for the short-range scheduling and intraday horizons where it actually drives a decision.

Forecast granularity questions

Should a contact centre forecast in 15-minute or 30-minute intervals?

30-minute intervals are the default and the right choice for most voice operations. Use 15-minute intervals when volume is high enough that each 15-minute bucket carries roughly 30+ contacts, demand is volatile within the hour, and the operation manages/schedules at 15-minute resolution. Avoid 15-minute intervals at low volume — random arrival variation dominates, the forecast becomes noise, and the Erlang C requirement becomes unstable and tends to over-staff. 60-minute intervals are too coarse for live voice (they average away the within-hour peak that drives the staffing need) but are appropriate for email, back-office, and low-volume deferred work. Match the interval to the volume per interval and to how fast the operation can act on the information.