Most shift businesses track their no-show rate. Far fewer have ever put a number on what one costs.
That gap matters, because the rate on its own is not decision-useful. Eight percent sounds tolerable. Eight percent of 200 shifts a week, at what it really costs you to cover each one, usually does not.
Here is how to work out your own number. No benchmarks, no industry averages, because those are mostly invented anyway and yours will be different.
The four numbers you need
1. Shifts filled per week. Not headcount. The number of individual shifts you actually staff.
2. Your no-show rate. If you don't track it, take the last four weeks of dispatch records and count the shifts where the assigned worker did not turn up and did not cancel in time to be replaced cleanly. Late cancellations that you covered without scrambling are a different, cheaper problem.
3. What a scramble costs. This is the one most businesses underestimate, because most of it is staff time rather than an invoice. When someone doesn't show, somebody spends the next thirty to ninety minutes calling down a list. Add the premium you pay to whoever picks it up at short notice. Add the overtime if the shift runs long because the replacement arrived late.
If you have never measured this, ask the person who actually makes those calls. They will have a better estimate than you do.
4. The share of no-show shifts that go unfilled. Not every scramble works. When it doesn't, you lose the revenue on that shift, and depending on your contracts you may owe a credit on top.
Putting it together
The arithmetic is deliberately simple:
no-shows per week = shifts per week × no-show rate
cost per no-show = scramble cost + (share unfilled × revenue per shift)
annual cost = no-shows per week × cost per no-show × 52Two things people get wrong here.
They count only the unfilled shifts. The covered ones are not free. They cost you the scramble, and they cost you the goodwill of whoever picked up the phone at 6am. That is a real cost, and it shows up later as that person becoming harder to reach.
They use an annual figure they never revisit. Your no-show rate is not a constant. It moves with the season, with which clients you're staffing, and with how well you know the people on your list.
Why the rate is so hard to move
Most attempts to reduce no-shows fall into three buckets, and all three have the same weakness.
Penalties. Deactivate after two no-shows. This works, sort of, and it costs you supply. In a tight labor market you are removing people faster than you can replace them, and the ones who leave first are often the ones with options, which is not the same set as the ones who don't turn up.
Incentives. Bonuses for attendance streaks. These work while you pay for them. The behavior usually does not survive the bonus ending.
Better communication. Confirmation texts, reminders the night before. Genuinely worth doing, and the cheapest thing on this list. But it moves the forgetful no-shows, not the deliberate ones.
The common weakness: all three treat every worker as an unknown until they have already let you down. At dispatch time, the moment the decision actually gets made, you usually cannot tell the person who has turned up for forty consecutive shifts from the person who is on their second week and has already dropped one.
Your best schedulers can. They carry it in their heads, or in a private spreadsheet, or in the shorthand they use when they talk about people. That knowledge is real, it is valuable, and it is completely trapped: it does not survive that person leaving, it does not transfer between branches, and it is invisible to anyone reviewing the roster.
What actually helps
The useful move is unglamorous: make reliability visible at the moment you assign the shift.
That can be as simple as a column in your scheduling sheet: shifts assigned, shifts turned up for, over a rolling window. Keep it factual and keep it recent. Six-month-old data mostly measures who has been around longest.
Two rules if you build this yourself:
Count the outcome, not the excuse. Reasons are worth recording for a conversation. They are not worth building a score out of, because everyone's reasons are reasonable and you'll end up adjudicating.
Show the person their own record. People dispute numbers they cannot see, and comply with numbers they can. It also catches your own data-entry errors, which there will be more of than you expect.
That gets you most of the way. Its limit is that it stops at your own front door. A worker who has been flawless for you for two years arrives at the next agency as a blank sheet, and someone with a bad record at three other agencies arrives at yours the same way.