Buffer times: why the diary is full and the day still does not work

Every business that gives out appointments knows the experience: the diary looks good in the morning, by eleven you are ten minutes behind, by four o'clock half an hour, and the last appointment of the day ends long after closing time. The next day the same, although nobody dawdled.

The usual explanation is that you need to plan better. The right explanation is less comfortable: a day with no air in it cannot come out right, because lateness adds up and being ahead evaporates. This article explains why that is so, and what follows from it for the way you give out appointments.

Appointments never take exactly as long as planned. Some go faster, some slower — so far, no insight. What matters is what happens to each of the two cases.

An appointment that takes ten minutes longer pushes the next one back by ten minutes. The one after that also starts late, and so on until closing. The lateness travels through the whole day.

An appointment that was ten minutes shorter pushes nothing. The next person arrives at the agreed time, not ten minutes early. The ten minutes you gained are gone.

That is the whole mechanism. Overruns accumulate, undershoots evaporate. So even if the durations are exactly right on average, the day still drifts systematically backwards. It takes no bad planning to produce that; it is enough that the actual duration varies.

From which follows the point that cannot be argued away: the only place where a day can shed lateness is a slot with no appointment in it. If there is none, there is no recovery. Buffer time is therefore not a comfort but the mechanism by which the plan corrects itself.

The spread is not a sign of disorder, it is the normal condition of any work on people and things. Four recurring sources:

• The thing itself. Two vehicles of the same model, two heads getting the same service, two consultations on the same topic: the effort varies, and it varies considerably.
• The beginning. Arriving, taking a coat off, greeting, a short discussion. Those minutes are rarely planned for and occur at every appointment.
• The in-between. Tidying, preparing, writing up, taking payment. That belongs to appointment time too, but often does not appear in the diary.
• The interruption. Phone, a query, a delivery. What comes from outside does not arrange itself around your grid.

The second and third points are the quiet cost drivers, because they are systematically absent. If the diary says 30 minutes, the work itself takes 30 minutes, and greeting, tidying and taking payment need another ten between them, then every single appointment is planned too short — not sometimes, always.

So check first whether your stored durations mean the whole appointment or only the work in it. In many businesses that one question settles more than any buffer rule. It is one of the points that has to be answered anyway when moving to online booking.

Now the part queueing theory has described for decades and that can be said without a formula: the closer a system comes to full utilisation, the more sharply waiting time rises — and not evenly, but ever more steeply.

At middling utilisation every variation has room to even out; a longer appointment falls into a gap and is forgotten afterwards. The fuller the plan, the rarer such a gap becomes. At complete booking no gap comes at all, and then there is nothing to absorb the backlog. So the last few percentage points of utilisation cost disproportionately much punctuality.

A worked example with round numbers. An eight-hour working day has 480 minutes. Plan it seamlessly into 30-minute appointments and that is 16 appointments. If each of them runs on average just five minutes longer than planned, the day ends 80 minutes late, and the last customer waits accordingly.

Take 14 appointments instead and leave 60 minutes of buffer spread across the day. The backlog is then 14 times five, that is 70 minutes, of which the buffer absorbs 60: the day ends ten minutes late instead of eighty. The price is two appointments, which is an eighth less capacity.

Whether that trade is worth it is a genuine business decision and not a rhetorical question. Two appointments fewer a day is noticeable. So are eighty minutes of lateness, unpaid overtime and people waiting — and those cost reputation and nerves on top.

Spreading buffer evenly between all appointments is the worst option: it is everywhere too small to absorb anything, and expensive in total. Put it instead where it does some work.

Two anchor points a day — one before the lunch break, one late in the afternoon — do more in practice than the same amount of buffer in small portions. They split the day into two sections that do not infect each other.

Keep the buffer visible to other people as well. A slot that looks empty in the diary will otherwise be filled with an appointment by the next person who opens it — and then it is no longer a buffer. Enter it as a block, with a name. In appointment systems like Salon Wizard's salon software that can be stored per service and per day instead of being kept free by hand each morning.

Once you carry those buffers into a week's plan, it becomes obvious whether the week still adds up: our rota template with an hours calculator totals the hours per person and per day as you enter the shifts. It runs in the browser and stores nothing.

Do not guess, measure. Two weeks are enough, and you need only two numbers per day: how many minutes behind the plan were you around lunchtime, and how many at the last appointment? Those two values say more than any utilisation statistic.

Then the reading. A backlog that builds up by lunchtime and grows in the afternoon calls for an anchor point before lunch. A backlog that only appears in the afternoon points at a particular service or at interruptions at a particular time of day. A backlog that is already there at the second appointment means a duration is simply stored wrongly.

If you note the actual durations per service at the same time, you can correct the planned values afterwards. Do not take the average for that, take a value that covers the longer cases too — an average guarantees by definition that every second appointment overruns.

One side effect is worth mentioning: if you have planned buffer in, a short-notice cancellation costs you less, because the gap joins up with the buffer into a usable block. Conversely, a loss hits a brim-full day harder. What else you can do about losses is set out in Why appointments fall through.