WFM guidePlanning foundations

Workforce planning assumptions

Q: What are the key assumptions in a workforce planning model?

The seven core planning assumptions in a contact centre WFM model are: (1) AHT (average handle time in seconds) — used directly as an input to Erlang C. A 10% AHT error creates approximately a 10% headcount error. (2) Shrinkage percentage — used to convert Erlang C required agents into scheduled agents. A shrinkage assumption that is 5pp below actual means every staffing plan is producing 5–8% fewer scheduled agents than needed. (3) Schedule adherence rate — determines how many scheduled agents are actually available. (4) Contact volume (by interval, day, week) — the primary driver of required agents. (5) Volume growth rate — determines how quickly headcount must be increased over the planning horizon. (6) Contact mix (channel split and contact type mix) — affects blended AHT and the channel-specific staffing model. (7) Service level target — the most direct driver of the Erlang C result; 5pp higher SL target can require 10–15% more agents at high volume.

Q: How often should workforce planning assumptions be reviewed?

Planning assumptions should be reviewed on three cadences: (1) Monthly — AHT, shrinkage, and adherence should be compared against actuals every month. If actuals diverge from assumptions by more than 5%, the assumption should be updated. Waiting for the quarterly review to correct a drifted AHT assumption can create a headcount error that compounds for three months. (2) Quarterly — volume growth rate, contact mix, and the full planning model should be validated against the quarterly actuals and revised for the next rolling quarter. The annual budget assumption set should be reviewed and updated. (3) Triggered reviews — any significant operational change (new product launch, regulatory change, technology implementation, major campaign) should trigger an immediate assumption review, regardless of the regular cadence. Planning assumption staleness is the most common cause of unexplained headcount gaps in contact centres.

Every headcount model is only as good as the assumptions that feed it. A 10% AHT error creates a 10% headcount error. A shrinkage assumption that is 5pp below actual systematically understaffs every shift. When two or three assumptions drift simultaneously, errors compound — and by the time the SL data reveals the gap, the operation has been understaffed for weeks. Documenting, validating, and maintaining planning assumptions is not administrative overhead — it is the foundation of every number in the staffing model.

The seven core planning assumptions

Average Handle Time (AHT)

300–600s for standard service; 700–1,200s for complex advisory/claims

Seconds per contact = (talk time + hold time + after-call work)

How it enters the model

Direct Erlang C input. AHT drives the traffic intensity (traffic = volume × AHT/3600). Higher AHT = more servers required to achieve the same SL.

Error impact

Linear. 10% AHT underestimate → ~10% headcount underestimate. AHT overestimate → overstaffing. The most common error is using blended AHT (including outlier long contacts) rather than the median/mode — blended AHT is pulled high by 1–2% of very long contacts.

Validation source

ACD contact history report. Weekly actuals vs. assumption. Segment by channel, contact type, and agent tenure to identify mix effects.

Shrinkage

25–35% total. Typical breakdown: absence 6–8%, training 4–6%, breaks 10–12%, meetings/admin 5–8%

% of scheduled time agents are unavailable to take contacts (absence + training + breaks + meetings + admin)

How it enters the model

Converts Erlang C required agents into scheduled headcount. Scheduled = required ÷ (1 − shrinkage). A 30% shrinkage means scheduling 43% more agents than Erlang C requires.

Error impact

If actual shrinkage is 32% but planned at 27%, each Erlang C output underestimates required scheduled headcount by ~7%. At 100 required agents, this means 137 scheduled instead of 143 needed — 6 agents permanently understaffed.

Validation source

WFM adherence reports, absence management system, training records. Break each component separately. Do not validate only total shrinkage — a low absence quarter can mask growing training shrinkage.

Schedule adherence rate

88–93% target. Below 85% indicates intraday management or culture problem.

% of scheduled time agents are actually in their scheduled activity (available when scheduled available, in break when scheduled break etc.)

How it enters the model

Determines actual agent availability vs. scheduled. At 90% adherence, a 100-scheduled-agent hour delivers 90 effective agent-hours. The headcount model assumes adherence at the target rate — if actual adherence is below target, effective capacity is lower than planned.

Error impact

A 5pp adherence gap (90% planned vs. 85% actual) means effective available agent hours are 5.5% lower than planned. At constant volume, SL falls proportionally.

Validation source

WFM real-time adherence report. Monthly adherence % by team. Identify teams consistently below target — these are intraday management or culture issues, not WFM model issues.

Contact volume (weekly, daily, by interval)

Varies by operation. Forecast accuracy (WAPE) target: ±5% at weekly level, ±15% at interval level

Expected number of contacts in each planning period. Decomposed into: weekly total, day-of-week distribution, intraday interval distribution

How it enters the model

The primary driver of required agents. Volume drives the Erlang C inputs per interval. Underforecasting volume = understaffing; overforecasting = overstaffing (waste).

Error impact

Non-linear at small scale (due to Erlang C's queueing behaviour). A 10% volume overforecast at a large operation has a near-linear 10% impact on headcount. At a small operation (10 agents), the same 10% overforecast may require 15% more agents due to the SL cliff effect.

Validation source

ACD reports. Weekly WAPE calculation: |actual − forecast| ÷ actual, averaged across intervals. Monthly WAPE report by interval to identify systematic error patterns (e.g., Monday consistently underforecast).

Volume growth rate

−5% to +20% annual for most operations. Influenced by customer base growth, product changes, IVR containment improvements, and channel shift

Expected % increase (or decrease) in total contact volume over the planning horizon

How it enters the model

Used to project future headcount requirements. Applied to the current volume baseline: Q3 headcount = Q2 headcount × (1 + quarterly growth rate). Multi-year capacity plans compound this assumption.

Error impact

Compounding. A growth rate assumption of 5% when actual is 10% creates a 5% shortfall in year 1, a 10% shortfall in year 2, and a 16% shortfall in year 3. Growth assumptions must be reviewed against actual volume trends quarterly.

Validation source

YoY volume comparison by month. Rolling 12-week growth rate vs. planned rate. Update the annual plan at mid-year with actuals to date.

Contact mix (channel and type)

Varies by operation. Monitor for mix drift — a 5pp shift from voice to chat changes blended AHT and the channel-specific staffing model.

The proportion of total contacts by channel (voice/chat/email) and by contact type (sales/service/complaint/technical)

How it enters the model

Determines blended AHT (used in Erlang C) and whether channel-specific models (concurrency for chat, backlog for email) must be applied separately. Mix drift changes the AHT assumption without any underlying change in individual contact handling time.

Error impact

Mix-driven AHT error is the most commonly missed planning risk. If voice contacts shift 10pp to chat and the planner uses last year's blended AHT, the model overstates voice AHT and understates chat throughput. Headcount gap is invisible until SL fails.

Validation source

ACD routing data by channel and contact type. Monthly channel mix report. Flag any month-on-month mix shift above 3pp for assumption review.

Service level target

Voice: 80/20 (industry standard) to 95/15 (high-performance). Chat: 80/30 to 80/60. Email: 90% within 4 hours is common for urgency-tier 1.

The SL target used as the Erlang C constraint: % of contacts answered within X seconds. For voice: typically 80% in 20s or 80% in 10s. For chat: response within 30–60s.

How it enters the model

The SL target directly determines the Erlang C required-agent output. The last 10pp of SL improvement is the most expensive — going from 80% to 90% SL in 20s may require 15–20% more agents. This makes the SL assumption the highest-impact assumption in the model.

Error impact

5pp SL target increase (80% to 85% in 20s) = 5–10% headcount increase. Operations that set aspirational SL targets without costing them create phantom budget gaps when they try to staff to target.

Validation source

ACD SL reports. Monthly actual SL vs. target. If actual SL consistently exceeds target, the target may be too conservative (potential overstaffing). If actual consistently falls short, the staffing model is understaffed.

Assumption validation cadence

Assumption	Review cadence	Update trigger	Owner
AHT	Monthly	Actual diverges from assumption by >5% for 2 consecutive weeks	WFM analyst
Shrinkage (total and components)	Monthly	Any component diverges by >2pp from assumption	WFM analyst + HR (absence)
Schedule adherence	Monthly	Monthly actual below target threshold (typically 88%)	Operations manager
Contact volume (interval level)	Weekly WAPE review	WAPE >10% at weekly level triggers model recalibration	WFM analyst
Volume growth rate	Quarterly	YoY growth rate diverges from assumption by >3pp in 2 consecutive months	WFM analyst + senior leadership
Contact mix	Monthly	Channel or type mix shifts >3pp month-on-month	WFM analyst + Ops
Service level target	Annual (budget) + triggered	Business decision to change SL commitment (contract, regulatory, or commercial)	Senior leadership + WFM lead

Planning assumptions questions

What are the key assumptions in a workforce planning model?

The seven core planning assumptions: (1) AHT — direct Erlang C input, 10% error = ~10% headcount error; (2) Shrinkage — converts required to scheduled agents; (3) Adherence — actual vs. scheduled availability; (4) Contact volume — primary demand driver; (5) Volume growth rate — forward-looking projection; (6) Contact mix — drives blended AHT; (7) Service level target — the most direct driver of Erlang C output, 5pp SL increase = 5–10% more agents.

How often should workforce planning assumptions be reviewed?

Monthly: AHT, shrinkage, adherence, volume (WAPE). Quarterly: growth rate, contact mix, full model validation. Triggered: any significant operational change (product launch, regulatory change, technology implementation). Assumption staleness is the most common cause of unexplained headcount gaps.