Planning for emails — the channel everyone under-plans
Email is the most under-planned channel in the contact centre
Most operations forecast voice in 15-minute intervals and plan staffing against an SL. Most operations forecast email as a daily total and staff it as a residual of voice — the agents not currently on a call do some emails. The first approach produces a measurable, manageable channel; the second approach produces a backlog that quietly grows until somebody escalates it. Email is treated as “voice with extra time” and it isn’t. It has a different arrival pattern, a different handle-time distribution, a different SLA structure, a different real-time playbook, and a different quality model. The fix is to plan it as a channel in its own right.
Why email is different from voice
Three structural differences shape every planning decision.
1. The arrival pattern is different. Voice arrives within a 90-minute window of when the customer needed help. Email arrives in a much wider distribution — customers send emails at 11pm, at 6am, from holiday, while travelling. The intraday curve of arrival is flatter and longer than voice. The daily curve is similar to voice but extends earlier and later.
2. The SLA is asynchronous. Voice SLA is “answered within 20 seconds.” Email SLA is “resolved within 4 working hours” or “within 1 working day” or “within 24 hours.” The customer doesn’t expect an instant answer; they expect a complete one. This is liberating — you can shape work into the day — and dangerous — you can put it off until the SLA breaks.
3. Email volume is not equal to email work. Voice has one call, one handler, one outcome. Email often has multiple touches per case: an initial reply, a customer response, a follow-up, sometimes a final close. The right unit of planning is the case, not the inbound email. Most operations measure inbound email volume, then wonder why their handle-time figures don’t add up.
The right forecast unit
For email, forecast in three layers:
Inbound new-case volume by day and intraday segment. Treat new-case arrival like voice arrival — daily curve, day-of-week effect, seasonality. New-case volume is the most predictable part of email.
Reply / follow-up volume. Each new case generates 0.5 to 2.5 additional touches on average, depending on how complex your work is and how good your first replies are. Track the ratio and forecast follow-up volume from it.
Backlog drain. Anything not handled today rolls into tomorrow. Today’s capacity isn’t just today’s arrivals; it’s today’s arrivals minus what gets carried, or plus yesterday’s carryover. The forecast has to model the backlog dynamic explicitly.
If you can’t answer the question “how many cases will be open at end of day tomorrow if I do nothing different?” your forecast isn’t complete.
Handle time — the number nobody trusts
Email handle time numbers in most systems are wrong. Three reasons: agents multi-task (working two emails in parallel undercounts both), the system captures keyboard time not thinking time, and a single email touch is not the same as a resolved case.
The clean way to measure: handler hours divided by cases resolved over a week. That gives you a defensible “hours per resolved case” figure that includes thinking, follow-ups, internal liaison, and the long tail of complicated cases. It will be higher than the system-reported “average handle time” and it will be the number that actually predicts your capacity need.
Staffing the channel properly
Email capacity is built differently from voice. There is no Erlang for email. The right model is:
Demand = forecast cases × resolved-hours-per-case + backlog drain target
Supply = handler hours × productivity factor
The productivity factor for email is usually 65–80% of paid hours — lower than voice productivity because email handlers are often blended, get pulled into other work, or absorb the “quiet time” gaps. Build it from your own historic data, not a textbook assumption.
Once you have demand and supply you can plan the channel like any other. Add cushion for the volatility of email volume (which is often higher than voice, day to day) and you have a defensible staffing model.
The operating model decision — dedicated or blended
The single biggest design choice for email is whether agents handle email as a primary role (dedicated email team) or alongside voice (blended).
Dedicated email teams usually deliver better quality, better SLA performance, and lower hours-per-case — the work is in flow rather than interrupted. They are harder to staff at low volume (you need a minimum critical mass) and tend to feel separated from the voice operation culturally.
Blended teams use email as fill-in work for voice gaps. Looks efficient on paper. In practice the queue logic almost always prioritises voice — agents work email only when voice is empty, which is rarely — and email runs into chronic backlog. Quality suffers because email is treated as a residual, not a craft.
The defensible answer for most operations of any size: dedicated email team, with a small blended skill pool to absorb voice peaks. Treat dedicated as the default and use blending as flex, not the other way round.
The real-time playbook for emails
Email real-time is unlike voice real-time. There’s no live queue depth in seconds; there’s a backlog in cases. The playbook:
Daily start-of-day backlog target. Real-time begins the day knowing how many cases are open and how many should be by end of day. This is the planning baseline. If the start-of-day number is already off, the day starts in deficit.
Mid-morning SLA breach check. Anything in the inbox already over its target response time is the highest priority. Real-time should pull SLA-at-risk cases to the front of the work queue.
Hourly aging report. Cases broken down by age — less than 2 hours, 2–4 hours, 4–24 hours, 1–3 days, over 3 days. The shape of that distribution tells you whether the operation is healthy or building backlog.
Backlog-spike playbook. When the backlog spikes (storm event, system outage triggering customer queries, marketing campaign), real-time should have a pre-agreed escalation: pull blended agents, offer overtime, defer non-urgent work, brief leadership. Don’t wait for the SLA to break.
The quality model
Email QA has to score what matters: did the response resolve the issue, was it accurate, was it well-written, did it close the case in one go? Most email QA scores formatting and signature lines — easy to score, irrelevant to customers.
The single most useful email-specific metric: first-touch resolution rate. The percentage of cases that were fully resolved with the first reply, without further customer back-and-forth. Operations that target FTR aggressively cut their total email volume by 15–30% within a quarter because they stop generating their own follow-up traffic.
Pair FTR with case-aging and you have a quality model that ties directly to operational performance. See designing a meaningful QA programme for how to build the form.
Common pitfalls
Treating email as voice with extra time. It isn’t. Voice answer-time doesn’t translate; the SLA is different, the work is different, the staffing model is different.
Forecasting inbound only. Inbound emails are about 40–60% of email work. If you forecast inbound and ignore follow-ups, you under-staff by 50% or more.
Ignoring the backlog. Today’s capacity has to include yesterday’s carryover. A forecast that doesn’t carry yesterday’s number forward is a forecast that hides the problem.
Blending as the default. Email-as-fill-in produces email-as-backlog. Most operations need a dedicated team with a blended flex layer, not the other way around.
Quality scoring formatting instead of resolution. If your QA form has 10 items and 6 of them are about layout, you have a form that produces compliant emails that don’t solve customer problems.
Conclusion
Email is a planned channel, not a residual one. Forecast inbound volume and follow-up volume separately. Measure hours-per-resolved-case, not system-reported handle time. Run a dedicated email team with a blended flex layer. Manage real-time by backlog and aging, not by SL-in-seconds. Score quality on first-touch resolution and outcome, not on signature lines. Do those things and the channel stops being the one that quietly breaks every quarter.
Pair with planning for multi-channel, designing a meaningful QA programme, real-time playbooks, and forecasting non-voice channels.