Skip to main content
TurnellaBeta
WFM guideForecasting

Contact centre volume patterns

Volume patterns are not just a forecasting input — they are the shape of demand that determines how the schedule should be designed. An operation that ignores pattern analysis schedules a flat headcount across the day — overstaffed in quiet periods and understaffed at the Monday morning peak.

The four layers of volume pattern

A complete volume model decomposes contact arrivals into four distinct layers. Each layer operates at a different time scale and is modelled and updated independently.

Intraday profile (half-hour intervals)

How contacts are distributed across the hours of the operating day. Most voice contact centres show a morning peak (09:30–11:30), a post-lunch recovery (13:00–14:30), and an afternoon secondary peak (15:30–17:30) before a decline towards close.

Typical range

Peak intervals typically receive 2.5–4× the volume of the quietest intervals in the same day.

Modelling approach

Calculate the proportion of daily volume that arrives in each 30-minute interval, averaged across the last 4–8 weeks of the same day type. Apply to the daily volume forecast to get interval-level estimates.

Update frequency

Review quarterly. Update immediately if a new service, channel, or promotion changes the intraday shape.

Day-of-week distribution

How volume is distributed across the days of the week. Most contact centres receive more volume on Mondays (post-weekend pent-up demand) and Fridays (end-of-week urgency) and less on Saturdays and Sundays where operating hours are reduced.

Typical range

Monday index typically 1.2–1.4× weekly average. Friday 1.0–1.2×. Weekend operations typically 0.5–0.8× weekday average.

Modelling approach

Calculate each day's average volume as a proportion of the weekly average (the day-of-week index). Apply to the weekly volume forecast to get day-level estimates.

Update frequency

Review monthly using a rolling 12-week average. Replace the 12-week average with the most recent 4 weeks if there is evidence of a structural shift.

Weekly seasonality

How weekly total volume varies across the year. Most operations show predictable peaks (Christmas, year-end, spring renewal periods) and troughs (summer holiday period, bank holiday weeks).

Typical range

Peak weeks typically 1.3–1.8× annual weekly average. Trough weeks (e.g. Christmas week if operating) typically 0.5–0.7× annual average.

Modelling approach

Calculate a seasonal index for each week of the year based on 2–3 years of historical data. Apply to the annual volume trend forecast to get week-level estimates.

Update frequency

Rebuild annually from the most recent 2–3 years of data. Update immediately after any event that creates a new seasonal peak (regulatory change, new product launch with annual billing cycle).

Trend

The underlying direction of volume growth or decline over time, independent of seasonal variation. A growing customer base, new product, or market expansion will show a positive trend. A self-service deflection programme will show a negative trend in the contact types it deflects.

Typical range

Trend rates vary enormously — from flat (0%) to high growth (10%+ per year) or decline (−5% to −20% for operations successfully deflecting to self-service).

Modelling approach

Fit a trend line (linear or exponential) to the de-seasonalised weekly volume data. The slope of the trend line gives the trend rate. Review whether the trend is accelerating, decelerating, or reversing.

Update frequency

Review monthly. If the trend rate has changed significantly (e.g. self-service deflection is working faster than expected), update the trend component to reflect the current rate rather than the historical average.

Detecting when a volume pattern has changed

Volume patterns are not static — they change as customer behaviour evolves, products change, and channels shift. The forecasting model must detect these changes before they produce sustained forecast errors. Three types of pattern change require different responses.

Sudden structural shift

Example

Self-service deflection goes live and inbound voice volume drops 15% in a single week

Detection signal

Single-week WAPE significantly above tolerance (>25% at daily level) in the same direction for 3+ consecutive days without an obvious event cause

Model update

Do not wait for the rolling average to adjust automatically — manually update the volume level assumption. If the shift is confirmed after 2 weeks of data, rebuild the trend component.

Gradual pattern drift

Example

Monday peak gradually shifts to Tuesday over 6 months as a competitor changes a billing cycle date

Detection signal

Monthly review of day-of-week indices shows Monday index declining from 1.35 to 1.15 over 6 months, Tuesday rising correspondingly

Model update

Update day-of-week indices from the most recent 8 weeks of data rather than the standard 12-week average. The more recent data better reflects the current pattern.

Seasonal pattern change

Example

A regulatory change creates a new peak in February that did not exist in prior years

Detection signal

Current-year February WAPE is persistently above tolerance. Prior-year February showed normal WAPE — so the error is not a persistent model problem, it is a genuine seasonal pattern change

Model update

Add a new seasonal adjustment for February in the model. Update the seasonal index for February based on the most recent 2 years of data, giving higher weight to the current year.

Volume pattern questions

What is a day-of-week index in contact centre forecasting?

A day-of-week index is a multiplier expressing how each day compares to the weekly average. If the weekly average is 1,000 contacts per day and Monday typically receives 1,300, the Monday index is 1.3. If Friday receives 800, the Friday index is 0.8. Indices are applied to the weekly volume forecast to distribute it across individual days. They are calculated from 12 weeks of historical data and should be updated monthly — day-of-week patterns can drift as customer behaviour changes.

Related guides