I have been offered a real-time analyst role with no planning background. How much maths do I need?

Less than the job title suggests and more than most people arrive with. You need to be fluent in five ideas: erlangs, occupancy, shrinkage, adherence and the service level curve. Every one of them is a calculator on this site with its working shown; an evening with each is enough to hold your own in the interview and the first month.

What distinguishes a good answer from a memorised one?

A number and a caveat. Anyone can say shrinkage is 30 per cent. A good answer says your last centre ran at 33, that leave and sickness were 16 of it, and that the figure was measured against paid time. Interviewers are listening for evidence you have seen the numbers move.

WFM analyst interview questions, with worked answers

Workforce planning interviews are unusually predictable, because the job runs on a small set of calculations and every interviewer wants to know whether you can do them without a tool. Here are the questions that come up, the answer that lands, and where on this site to practise each one. The numbers all use the same reference interval, so you can hold one example in your head: 200 contacts in half an hour, a five-minute handle time, an 80/20 target.

1. What is an erlang?

The average number of contacts in progress at once: arrival rate multiplied by handle time in the same units. 200 contacts in 30 minutes at 300 seconds is 6.67 a minute times 5 minutes, which is 33.3 erlangs. Say the number and the unit; interviewers hear “33 agents” often and it is wrong.

2. Why do we need more agents than erlangs?

Because arrivals are random and bunch. With exactly 33 agents the queue never clears; 39 agents hit 80/20 for that interval. The six extra are the price of the service level, and the price falls as a share for bigger queues, which is question 8.

3. What is the difference between service level and ASA?

Service level is the share answered within a threshold; ASA is the average wait across everyone including those who waited nothing. At 38 agents the interval delivers 75.6 per cent service level and a 21-second ASA. Both are right, and the second sounds fine while the first is a miss, which is why you manage to service level. Prove it in the service level calculator.

4. What is shrinkage, and what is a typical figure?

Paid time not available to the queue, split into external (leave, public holidays, sickness) and internal (breaks, training, coaching, meetings). 30 to 35 per cent is typical; under 20 usually means breaks were left out. The follow-up is always “how do you apply it”: divide the requirement by one minus shrinkage, and round up. The shrinkage calculator shows the split.

5. What is occupancy and what should it be?

Handling time divided by available time. 85 per cent is the usual ceiling for voice; above 90 for long stretches burns agents out and raises absence, which raises shrinkage, which raises occupancy. Mention that small queues need lower occupancy for the same service level. The occupancy calculator has the numbers.

6. Occupancy, utilisation, adherence: what is the difference?

Three denominators. Occupancy is handling over available time; utilisation is handling over paid time, so it includes shrinkage; adherence is time in the scheduled state over scheduled time. An agent can be 85 per cent occupied, 55 per cent utilised and 94 per cent adherent at once. The adherence calculator and the utilisation calculator each show one pair.

7. How do you get from a monthly volume to an FTE figure?

Month to week (× 12 ÷ 52), week to days by the weekday split, day to intervals by the intraday shape, Erlang C per interval, occupancy ceiling, shrinkage, sum the agent-hours, divide by contracted hours, divide by roster efficiency. Say the chain in order and give the surprising number: the answer is roughly double the naive workload figure. The FTE guide walks it.

8. Why do small teams cost more per contact?

Pooling. Four queues of ten erlangs need 56 agents between them; one queue of forty needs 46. Every small queue carries its own slack. This is the answer to “should we multi-skill” and it has a calculator: pooling benefit.

9. What is a good forecast accuracy, and how do you measure it?

At interval level, a WAPE under 10 per cent is good. Say WAPE rather than MAPE and explain why: WAPE weights by volume, so a big miss on a quiet interval does not dominate. Then say which forecast you measure against, the one the roster was built on. The forecast accuracy calculator computes both.

10. It is 11 a.m. and the morning ran ten per cent hot. What do you do?

Reforecast the rest of the day at the observed ratio, damped, rerun the requirement for the afternoon peak, and move breaks and coaching out of the intervals that are short before asking for overtime. This is the real-time analyst’s whole job in one sentence; the intraday reforecast calculator does the sum.

11. How would you tell whether a roster is any good?

Not by coverage, which is total rostered hours over total required. By fit, which compares each interval and penalises both surplus and shortfall. A roster can be 102 per cent covered and 91 per cent fit, with the shortfall all in the afternoon. The schedule efficiency calculator is the demonstration.

12. What is the one mistake you see most in spreadsheet staffing models?

Applying shrinkage twice: once per interval and again by dividing by productive hours at the end. It overstates the requirement by the shrinkage multiplier, which at 30 per cent is 43 per cent. Interviewers who have built the spreadsheet will smile.

Practise the reference interval

Inputs

Results

Agents required

On the phones for the whole interval, before shrinkage.

Service level achieved: 82.6%
Occupancy: 85.5%
Average speed of answer: 13.4 s
Contacts that will wait: 25.4%
Workload: 33.3 erlangs

Show the working

Workload = 200 contacts ÷ 30 min × 300 s ÷ 60 = 33.3 erlangs.
The queue is only stable with more agents than erlangs, so the search starts at 34.
With 39 agents, Erlang C gives a 25.4% chance that a new contact finds every agent busy.
Service level = 1 − 0.254 × e^(−(39 − 33.3) × 20 ÷ 300) = 82.6%, which clears the 80% target.
With 38 agents it would be 75.6%, which does not.
Occupancy = 33.3 ÷ 39 = 85.5%.

Doing this for every interval of the week? Pebble WFM computes the requirement from your forecast and builds the roster. Free month, no card needed.

Where next

How to calculate Erlang C staffing: questions 1 and 2 in full.
When a spreadsheet stops being enough: the answer to “what would you change about our planning process”.