How to Estimate Wait Times Accurately (Without Guessing)

"About ten minutes" is the most expensive sentence at your front desk. Say it and be wrong, and you have converted a patient customer into an annoyed one. Here is how to produce an estimate you can actually stand behind — using arithmetic you can do on the back of a receipt.

The one formula you need

For a single queue with a single server, the estimate is:

Six people ahead, two counters open, four minutes per customer: (6 ÷ 2) × 4 = 12 minutes. That is it. Everything else in this article is about getting those three numbers right and knowing when the formula lies to you.

Measuring your real service time

Almost nobody knows their true service time, and almost everybody thinks they do. The guess is nearly always too low, because you remember the smooth transactions and not the one where the card machine failed.

Measure it properly once and you are set:

1. For one representative week, record the time each customer is called and the time they are finished.
2. Take the difference for each, and put them in order.
3. Use the median, not the mean. One 45-minute complaint drags a mean upward and makes every estimate too pessimistic.

A queue system that timestamps status changes gives you this data as a by-product, which is the main reason to prefer measured estimates over a number somebody typed into a settings box in 2023.

Segment before you average

One average across genuinely different jobs produces an estimate that is wrong for everyone. A pharmacy counter where a repeat prescription takes 90 seconds and a consultation takes 12 minutes has no meaningful "average customer".

If two kinds of work differ by more than roughly a factor of two, split them into separate lines with separate service times. You get better estimates and a fairer order, because the quick job stops sitting behind the slow one.

Pad it, but pad it honestly

Under-promise and over-deliver is good advice that is easy to overdo. Two guidelines:

• Quote a number a bit above the median — the 60th to 75th percentile of your measured times works well. Most customers then beat their estimate, and the ones who do not are not badly out.
• Do not pad by more than about 25%. Quote 20 minutes for a 10-minute wait and customers leave, or they learn your estimates are fiction and stop believing them. An estimate is a reputation you spend.

Ranges help where you are genuinely unsure: "10–15 minutes" is more honest than a false precision of "12 minutes", and customers read the top of the range anyway.

Update the estimate, do not just issue it

A wait estimate printed once at the door is a guess about the future that immediately starts rotting. The thing that actually keeps customers calm is an estimate that moves: 14 minutes, then 11, then 7. Watching it count down is a progress bar, and progress bars are what make waiting tolerable.

This is the practical reason to put the ticket on the customer's phone rather than on paper — paper cannot revise itself. It is also why you should recalculate when a counter opens or closes, not only when someone is served.

When the formula lies

Know the failure modes, because they are the times your estimate embarrasses you:

• Near full capacity. When customers arrive about as fast as you serve them, waits stop rising gently and start exploding. At 90% utilisation a small surge produces a wait far longer than the arithmetic suggests. If you are regularly at that point, no estimate will save you — you need another counter open.
• Staff breaks and shift changes. The formula assumes counters stay open. Feed the actual number of open counters into it, and recalculate the moment one closes.
• The start of the day. With three customers served, your rolling average is noise. Fall back to your historical median until you have a reasonable sample.
• Known-slow customers. If your staff can see that the person being served now is a complicated case, the queue behind them is understated. This is where a human override beats the maths.

When things go wrong, say so

A system down, a staff member off sick, a delivery blocking the counter. The instinct is to keep quoting the normal estimate and hope to catch up. Do the opposite: raise the estimate immediately and tell people why. Customers forgive a stated delay far more readily than one they discover by watching an estimate quietly fail to come true.