Metric Definitions

A complete reference for every metric, input, and formula used in Turnella.

Demand Metrics

Calls / sessions / items per interval

The primary demand input: how many contacts arrive in your planning unit. For voice and chat, this is per half-hour interval. For email and back-office, this is the daily total (because deferred work is planned over the day, not the interval).

Always size your inputs on your busiest representative interval, not a daily average. Planning to the average guarantees you are understaffed for half of your operating hours.

AHT — Average Handle Time

The mean time an agent spends per contact, measured in seconds.

• Voice AHT: talk time + after-call work (wrap-up). Must include both. Using talk time alone underestimates headcount.
• Chat AHT: end-to-end session duration from first message to resolution.
• Email AHT: processing time per item from opening to resolution.

AHT is the single most powerful driver of headcount — more powerful than volume. A 10% increase in AHT has roughly the same headcount impact as a 10% increase in volume.

Existing backlog

For email and back-office channels: the number of items already waiting in the queue at the start of the planning period. Clearing this backlog requires capacity above keeping up with new arrivals — the model adds FTE to clear it within your SLA deadline.

Service Target Metrics

Service level target

The percentage of calls you aim to answer within your target wait time — expressed as "X% in Y seconds". The most common standard in contact centres is 80% of calls answered within 20 seconds (written as 80/20).

Why 80/20? It was first adopted by AT&T in the 1980s as a rule of thumb based on empirical research showing that after 20 seconds, callers start to abandon. It has no scientific basis as a universal optimum — your customers and context should drive your target.

Target answer time

The "within X seconds" threshold in your SL target. Typical range: 10–30 seconds. For voice, 20 seconds is the most common standard. For chat, first-response SL targets are typically 30–60 seconds.

Max occupancy

A cap on how busy agents are kept. Occupancy is the percentage of time agents spend actively handling contacts (versus idle time between contacts). At 85% occupancy, agents have approximately 9 minutes of idle time per hour.

The max occupancy setting acts as a guardrail: if the Erlang C formula would result in occupancy above this threshold, Turnella adds extra agents until occupancy falls below it.

Why cap it? High occupancy creates a fragile queue. With no buffer, any small increase in volume or handle time cascades immediately into a queue build-up. Keeping occupancy below 85% gives the system a margin of safety.

Recommended setting: 85%. The typical healthy range is 70–80% average, with peaks touching the ceiling in busy intervals.

Shrinkage

The percentage of scheduled time during which agents are not available to handle contacts.

Concurrency (chat)

The number of simultaneous chat sessions one agent handles at once. Higher concurrency reduces headcount but increases response time per session and effective AHT. The recommended range is 2–4 sessions per agent.

Active time ratio (chat)

The fraction of a chat session during which the agent is actively engaged (typing, reading, processing). The remaining time is "wait time" where the customer is composing their message. Typical range: 0.60–0.75. Used in the concurrency model to estimate true agent utilisation per session.

Forecast Accuracy Metrics

WAPE — Weighted Absolute Percentage Error

The primary forecast accuracy metric for contact centre forecasting.

WAPE = Σ|Actual − Forecast| ÷ Σ Actual × 100

WAPE weights high-volume intervals more heavily than low-volume ones — a 50% error at 02:00 barely moves the score, while a 10% error at your peak afternoon slot has significant impact. This is the right behaviour for operational planning.

MAPE — Mean Absolute Percentage Error

The simple average of each interval's absolute error as a percentage of its actual value. Easier to explain than WAPE but easily distorted by very low-volume intervals (a 1 call actual vs a 2 call forecast = 100% error, which is technically correct but operationally irrelevant). Always use alongside WAPE, not instead of it.

Forecast bias

The systematic directional error in your forecast.

• Positive bias: you consistently over-forecast (risk of chronic overstaffing and wasted cost).
• Negative bias: you consistently under-forecast (risk of chronic understaffing and missed SL).

A well-calibrated model should have bias close to zero. Persistent bias in one direction means the model is systematically wrong and should be reviewed.

Within 5% and Within 20%

The percentage of forecast intervals that fell within 5% (or 20%) of the actual volume. ≥ 70% within 5% is a useful bar for high-confidence operational planning.

Cost Metrics

On-cost uplift

Employer costs on top of wages, expressed as a percentage of base wage. Turns a raw hourly rate into the true fully-loaded cost of an hour worked. UK typical: 22–28%.

Components: employer NI (~13.8%), employer pension (min 3%), holiday pay loading (~12%), benefits.

Loaded hourly rate

Base hourly rate × (1 + on-cost uplift ÷ 100). This is the true cost of one hour of work, including employer costs. Example: £15.00 base × 1.25 = £18.75 loaded.

OT premium

Overtime pay rate expressed as a percentage of the base hourly rate. 150 = time-and-a-half; 200 = double-time.

Cost per contact

Weekly labour cost ÷ weekly contact volume. The primary labour efficiency metric — allows comparison across periods with different volumes.

Rostering Metrics

FTE (Full-Time Equivalent)

A normalised headcount unit. 1.0 FTE = one full-time employee working 40 hours per week. A part-time agent on 20 contracted hours = 0.50 FTE.

FTE = Contracted hours per week ÷ 40

Contracted FTE vs Scheduled FTE

• Contracted FTE: total FTE from all active roster entries (what your payroll covers)
• Scheduled FTE: total FTE deployed in the current schedule plan (what your plan calls for)
• Variance: the gap between them. Positive variance = schedule requires more than contracted (overtime risk). Negative variance = agents are under-utilised relative to their contracts.

Bradford Factor

Disclaimer: The thresholds and responses below are illustrative examples only. Always follow your own HR policy and local employment law. Turnella does not provide HR or legal advice.

Used to assess the impact of employee absence patterns.

B = S² × D

S = number of separate absence spells in a 12-month rolling period
D = total days absent

The formula squares the spell count to penalise frequent short absences more severely than a single long illness. Three separate 1-day absences score 3² × 3 = 27. One 9-day absence scores 1² × 9 = 9. The frequent short-absence pattern (more disruptive to scheduling) scores three times higher.

Check your organisation's absence management policy for the specific thresholds it uses.

Productive Capacity

The number of productive hours an agent actually delivers in a year, after deducting all planned and unplanned absence from their contracted hours.

A 40h/week agent typically delivers 30–34 productive hours per week — 75–85% of contracted hours. This is why contracted FTE and required FTE are not the same number, and why headcount budgets built on contracted hours alone are almost always short.