Forecasting answers one question: how many people are needed in each time slot. Shift scheduling answers a harder one: which specific people will be available in those slots, with which days off, which breaks, and under which employment rules. A good part of the capacity an operation thought it had bought disappears between the two questions.
Scheduling often gets treated as paperwork that happens after planning. It is not. A well-sized operation that is badly scheduled performs like a smaller one: there are spare people when nothing arrives and too few when everything lands at once.
Sizing and scheduling are two different jobs
Sizing means working out how many people are needed to sustain a volume at an agreed service level. That calculation is in how to size a support operation. Scheduling means spreading the people you already have across the days and hours where the work actually is.
You can have exactly the right headcount and still miss the service level every Tuesday at ten in the morning. When that happens, hiring does not fix it — moving shifts does. Hiring to cover a scheduling gap is the most expensive way to close it.
The curve rules, and an average hides it
Work does not arrive evenly. It arrives concentrated in certain hours, certain weekdays and certain dates in the month, and that shape is fairly stable even when the absolute level moves. Which is why the planning unit is not the day: it is the interval, usually half an hour. A daily average produces a reassuring figure that describes no real moment of the day.
The day-of-week pattern is normally more stable than total volume, and the two are worth handling separately: the level gets forecast, the shape gets observed. How that base is built is covered in demand forecasting in BPO. And it has to be revisited whenever the business changes: a campaign, a shift in the billing date or a new channel all move the curve without warning.
Voice and back office are not scheduled the same way
In voice and chat, work is handled the moment it arrives or it is lost. The schedule has to track the curve interval by interval, because an uncovered hour cannot be recovered later.
In back office the work waits in a queue and what rules is the delivery commitment, not the arrival time. That gives room to level load across days, provided the commitment is written down and cases are worked oldest first. The difference between channels is laid out in support channels: voice, chat and email.
Mixing both logics into one schedule is a common mistake: the back office ends up absorbing people exactly when calls peak.
Constraints come before optimisation
Before looking for the most efficient schedule, write down what the schedule is not allowed to do. These constraints show up every time:
- Employment framework. The applicable working week, rest periods, overtime limits, and premiums for night, Sunday and public holiday work. This is set by local law and should be validated by the right professional; none of it is legal advice. How the working week feeds into cost is in the working week and cost per agent.
- Committed coverage. The hours the contract promises to cover, which do not always match the hours the work arrives in.
- Overlap with the client. The hours the client's own team is available to resolve escalations, covered in time zones and overlap.
- Commuting and safety. For shifts that start very early or end very late, transport is a real constraint, not a wellbeing detail.
- Offline activities. Training, coaching, meetings and calibration exist and consume hours. If they are not in the schedule, they come out of coverage.
Fixed, rotating and split shifts
Each pattern solves something and costs something:
- Fixed. The same hours every week. It is what people prefer and what helps retention most, but it fits badly against a sharply peaked curve.
- Rotating. Good coverage across morning, afternoon and night, and the hardest on people when the rotation is unpredictable. A published, stable cycle changes its effect entirely.
- Split. Fits a two-peak curve very well and is the hardest to sustain: the split time has to be added to the commute. It tends to work only when the setup is remote and explicitly accepted.
- Short top-up shifts. They cover one specific peak without disturbing the rest of the schedule. They need trained people who are available for that block.
The choice of pattern is not neutral for retention, and getting it wrong is paid for in recruiting: the real causes sit in attrition in BPO.
Being over and being under do not cost the same
An overstaffed interval costs the value of those hours. An understaffed interval costs queueing, abandonment, rework, pressure on the team and, where penalties apply, money. The asymmetry is large and it is almost always ignored when the schedule is built, because it gets optimised as if both deviations weighed the same.
A schedule that balances on the weekly total and balances in no single interval is not a schedule: it is a spreadsheet.
It is worth reviewing coverage interval by interval and deciding deliberately where you will be short, rather than discovering it on Tuesday at ten.
A schedule only counts if it is met
The gap between scheduled hours and hours actually online is what decides the outcome. If adherence is low, any improvement in scheduling evaporates before it reaches the operation. That deduction — holidays, absence, training, breaks, offline time — is explained in shrinkage and adherence, and it belongs in the plan rather than in the post-mortem.
How far ahead it gets published
How far ahead the schedule is published is a retention variable, not an administrative detail. Someone who knows their hours for the coming weeks can organise their life; someone who finds out on Friday afternoon cannot. Publishing on a stable horizon and honouring what has already been communicated cuts absence and resignations more than most wellbeing initiatives do.
Predictable peaks are handled by a different mechanism and planned separately, as described in seasonal peaks. Overtime is the last resort, not the default buffer: when it shows up every week, the problem is in the base schedule.
The agent who can handle everything does not exist
Scheduling is easier if you assume anyone can take any case. That assumption is rarely true: there are products, languages, systems and escalation tiers that only part of the team handles. The schedule has to balance by skill, not only by total headcount, and it helps to know how many certified people exist for each critical skill. An operation with one single person able to resolve a case type has a continuity problem, not a scheduling problem.
What to review every week
- Scheduled coverage against required coverage, interval by interval rather than in total.
- Actual coverage against scheduled: absence, late starts and offline time.
- Intervals that missed the service level, and whether the cause was volume or coverage.
- Last-minute shift changes and why they happened. If they repeat, the schedule is not realistic.
- Overtime per week, read as a symptom rather than as a solution.
That is enough to separate what needs fixing in the schedule from what needs fixing in the forecast, which are different problems with different answers.
How smartBPO works it
We plan by interval rather than by daily average, and we keep the shape of the curve separate from the expected volume level. We write the constraints down before building the schedule: applicable employment framework, committed coverage, overlap with the client's team, and offline activities. We schedule by skill, not only by headcount. We publish shifts ahead of time and hold to what was published, because a roster that changes without notice is one of the most avoidable reasons people leave. And we review coverage interval by interval with the client every week, so the conversation is about where the gaps are rather than about whether the total adds up.