Voice processes are staffed with queueing models because a customer is waiting on the line and every second counts. Email, ticket, claims and data-processing work is different. The item can wait in a queue for hours without anyone hanging up, as long as it is finished within an agreed turnaround time (TAT). That single difference changes how capacity is planned, and makes the method much more transparent.
Start with workload, not headcount
The core idea is simple: work arriving × time per item = workload hours. Divide by the productive hours one person delivers and you get the people needed.
Take an email support team for an online travel company receiving 3,000 emails a day. Not every email takes the same time, so the handling time is a weighted average of the main types:
| Email type | Share | Minutes each |
|---|---|---|
| Booking confirmation and simple queries | 60% | 3 |
| Changes and cancellations | 30% | 7 |
| Refund disputes needing investigation | 10% | 12 |
Weighted average: (0.6 × 3) + (0.3 × 7) + (0.1 × 12) = 5.1 minutes. Daily workload is 3,000 × 5.1 ÷ 60 = 255 hours.
The mix matters more than people expect. If a policy change pushes refund disputes from 10% to 20% of volume, at the expense of simple queries, the average rises to 6.0 minutes and workload grows by about 18% with no change in email count. Tracking volume alone would miss it.
Productive hours per person
An associate logged in for 8 hours does not process email for 8 hours. Breaks, team huddles, quality feedback sessions, system downtime and reading process updates all consume time. Suppose 80% of logged-in time is productive. That gives 6.4 hours per associate per day.
255 ÷ 6.4 = 39.8, so 40 associates need to be working on any given day.
Then come the people who are on the roster but not at work: planned leave, sick leave, unplanned absence. If these average 10%, the team needs to schedule 40 ÷ 0.9 = about 44.3, so 45 people per day. If the operation runs seven days a week and each associate works five, the headcount becomes roughly 45 × 7 ÷ 5 = 62 or 63. Many first-time planners stop at 40 and wonder why they are always short.
Turnaround time changes the question
In voice, the question is “how many agents are needed in this half-hour?” In non-voice, the question becomes “can all work that arrived be completed before its TAT expires?” A 24-hour TAT gives the team a lot of flexibility. Emails that arrive in a morning spike can be worked through the afternoon, so staffing can be much flatter across the day than the arrival pattern.
Tighter TATs remove that flexibility. With a 2-hour TAT, staffing has to follow arrivals almost as closely as a voice queue, and the planning starts to look more like interval-based scheduling. Chat sits at the far end of the spectrum: it behaves like voice with concurrency, because customers wait in real time.
TAT definitions vary, and they matter. Is it calendar hours or business hours? Does the clock pause while the team waits for information from the customer or from the client’s own department? Is it measured to first response or to resolution? A team can comfortably meet a 24-hour first-response TAT while taking five days to actually resolve cases, and the client’s customers will judge the second number.
The backlog is the honest scoreboard
Backlog is the work received but not yet completed. Because non-voice items wait instead of abandoning, under-capacity does not show up immediately. It accumulates.
Using the numbers above, 40 associates at 6.4 hours each provide 256 hours, enough for about 3,010 emails at 5.1 minutes. On a typical day that just covers arrivals. Now suppose Monday brings 3,600 emails because of weekend travel disruptions. About 590 emails are left at the end of the day. If Tuesday is a normal day, the team still starts behind and ends behind. Without overtime, extra staff or a quieter day, the backlog persists, and every one of those emails is ageing towards its TAT.
That is why non-voice MIS should show not only backlog count but backlog age: how many items are under 4 hours old, 4 to 12, 12 to 24, and beyond TAT. A backlog of 600 fresh emails is a normal morning. A backlog of 600 emails that are 20 hours old is a breach about to happen.
Work allocation affects this too. If associates pick their own items, they often pick the quick ones, leaving complex cases to age. Push-based allocation, or rules that prioritise items closest to breach, keep the age profile healthy.
Getting the time-per-item right
The whole model rests on the handling time per item. In voice, the ACD measures it automatically. In non-voice, it often comes from a time-and-motion study done at transition, then never revisited. Processes change: new fields in a form, an extra approval step, a new system that loads slowly. Each adds seconds that multiply across thousands of items.
Workflow tools that log when an item is opened and closed help, but raw durations include interruptions and items left open during a break. Trimming extreme values, or using the median per item type, gives a more stable standard. Refreshing standards every quarter, and after any process change, keeps the plan honest.
Productivity targets for associates are usually built from the same standards: if a refund dispute takes 12 minutes and productive time is 6.4 hours, the expectation is about 32 disputes a day. When targets are set from outdated standards, associates either cannot meet them or meet them easily, and in both cases the capacity plan is wrong in the same direction.
Written by the RP Strativa Team. Figures in the examples are illustrative.
Leave a Reply