Want help applying this guide?Ask the DripTell team
A support manager sees next Tuesday's forecast and gets one clean number. The team expects more conversations. That still does not answer the question people actually have to solve: who should be working at 10:30, on which queue, with enough room for breaks, training, absence, and the backlog already waiting?
A useful staffing plan translates expected demand into timed coverage. It does not divide a daily total by an average number of cases per agent. It works by queue and interval, keeps productive need separate from scheduled headcount, and records the assumptions a manager can change when reality moves.
Start with work that belongs together
Begin with a reviewed support volume forecast. Keep queues separate when they have different skills, service promises, handle times, concurrency, or operating hours. Combining email, live chat, and specialist cases into one daily total hides when the work arrives and who can handle it.
Choose the interval that matches the decision. Fifteen or thirty minutes can make sense for phone and live chat. A daily view may be enough for slower asynchronous work, but it still needs a starting backlog and an age or completion target.
Microsoft's current capacity planning guidance supports short term planning in 15 minute intervals and exposes capacity by channel and queue. The exact product is less important than the discipline: demand must be visible at the level where coverage changes.
Convert arrivals into productive need
For each queue and interval, combine expected arrivals with observed handling effort and the service promise. Voice usually needs a queueing model because work cannot be parked safely. Messaging may allow controlled concurrency. Email or tasks may be planned against completion time and backlog rather than seconds to answer.
- Comparable queuesKeep work with different skills, timing, or service promises separate.
- Visible assumptionsRecord effort, service targets, concurrency, backlog, and forecast version.
- Honest coverageAdd shrinkage, minimum skills, opening duties, rest, and leave.
- Workable shiftsInspect interval gaps instead of trusting the daily total of paid hours.
- Owned triggersName the action and owner for high and low demand scenarios.
Do not use one optimistic capacity number for every channel. The companion guide on safe conversation capacity explains why an agent handling two simple chats is not automatically able to handle two complex ones. Use a recent, comparable effort distribution and document any concurrency assumption.
At this stage the output is productive need, not the number of people to put on a roster. It describes how many people must be available for customer work if nothing else interrupts them.
| Planning input | What it changes | Evidence to keep |
|---|---|---|
| Arrivals by interval | Work entering the queue | Forecast version and event assumptions |
| Handling effort | Productive workload | Median and upper range for comparable work |
| Service target | How quickly coverage must absorb demand | Published target and eligible contacts |
| Concurrency | Simultaneous messaging load | Tested safe limit by work type |
| Starting backlog | Work already owed at opening | Count, age, and completion promise |
Add coverage that the forecast cannot supply
Now turn productive need into scheduled headcount. Add shrinkage for paid time when people are not available for customer work, including breaks, coaching, meetings, training, planned leave, and a defensible absence allowance. Keep those causes visible. One unexplained percentage is hard to improve and easy to weaponize.
Then apply the constraints the operation must respect. A queue may need a minimum trained person even when the statistical forecast rounds toward zero. Opening and closing duties need owners. Specialist cover may need an overlap period. Labor rules, contracts, rest, and accessibility needs limit which shifts are possible.
Microsoft's current capacity planning guidance treats service level, target answer time, shrinkage, concurrency, channel, and queue as explicit planning inputs. Keep forecast driven coverage separate from operating constraints such as minimum staff, working hours, and rest. That separation is useful even in a spreadsheet.
Build a roster people can actually work
A staffing requirement is not yet a schedule. Map named or role based shifts to the requirement, then inspect the gaps interval by interval. A plan can contain enough paid hours for the day and still fail during the first hour, lunch, or the evening handover.
Use service level targets as a customer promise, not a reason to compress every break. Preserve training and coaching instead of treating them as spare capacity that disappears whenever demand rises. If the base plan depends on permanent overtime or perfect attendance, it is already a high scenario pretending to be normal.
Publish three views. The base view covers the likely forecast. A high view names the trigger for flex time, cross skilled help, callbacks, or queue overflow. A low view protects useful work such as coaching, quality review, and knowledge maintenance. Decide those actions before the queue becomes noisy.
Compare the plan with the live day
During the day, compare forecast arrivals, actual arrivals, required productive capacity, scheduled capacity, and actual available capacity. These are different signals. A bad forecast should not be disguised as poor schedule adherence, and a sudden absence should not be blamed on the forecast.
Watch occupancy with customer outcomes. High occupancy for a short peak may be expected. High occupancy across most intervals means the plan has no recovery room. Low occupancy may be a forecast miss, a quiet interval between peaks, or necessary cover for a specialist queue.
After the week, keep a small variance log. Record whether the miss came from arrivals, effort, backlog, shrinkage, skill coverage, schedule fit, or execution. Change the relevant assumption only. Rebuilding the whole model after one difficult day destroys the evidence that would improve it.
With DripTell's shared inbox, a team can keep queue ownership and conversation context visible while managers review staffing assumptions. The staffing method still belongs to the operator. The software should make the evidence easier to see, not replace judgment.
Frequently Asked Questions
How far ahead should a support staffing plan look
Use a longer view for hiring and training, and a short interval view for rosters. Refresh the short plan when forecasts, launches, leave, or operating hours materially change.
Should backlog be added to the new forecast
Yes, for asynchronous queues. Keep starting backlog separate from new arrivals, include its age, and give it a completion target so old work is not hidden inside a daily total.
Is shrinkage the same as occupancy
No. Shrinkage is scheduled time unavailable for customer work. Occupancy describes how much available productive time is spent handling work. Mixing them can create both understaffing and unfair performance claims.
What should happen when actual demand exceeds the plan
Use the pre-agreed high scenario. Trigger flex cover, cross skilled help, callbacks, or controlled overflow at a visible threshold, then record the cause and review it after service stabilizes.
DripTell Editorial
Practical guidance reviewed by the DripTell product and customer workflow team.
See how DripTell checks product claims, uses primary sources and handles corrections.
Editorial and source policy