Headcount tells you nothing
An NEMT company is three seats. Drivers, who deliver the service. Dispatch, who decide what happens today. And the back office that bills and collects, which is the seat that turns completed trips into money. Every company running five vehicles has all three, whether or not it has three people.
That last part is where the trouble hides. In most small operations one person occupies two seats, often the owner, and nobody notices until that person is away. Seat coverage tells you everything.
So write the seats down and ask three questions of each one. Who does this today. Who else could. And what would it take for a third person to learn it.
The depth chart
A depth chart is one page: every seat, who sits in it, who is second, and how many weeks that second person would need before they were actually useful rather than theoretically available.
The first two columns fill in instantly. The third is where you stall, and that is the point of the exercise. A seat where the honest answer is “nobody” is a seat that can stop the company, and if it touches cash or compliance it is a bigger risk than any vehicle in your fleet. A van breaking down costs you a route. Losing the person who does your billing stops the money coming in until someone else has learned what they knew.
There is only one way to find out whether a backup is real, and it is not a conversation. Pick a seat, and have the backup work it for a full day this month while the usual person does something else. You will find out the difference between someone who thinks they could cover and someone who can, and the gap is always something specific: a portal login nobody shared, a facility that only takes calls from one person, a step in the billing run that exists purely in muscle memory.
This is not the question Vision asks. That one is about what breaks without the owner. This one is about what breaks without anybody else, and an owner can pass the first test while quietly failing this one.
Recruit before someone quits
Drivers turn over in this industry. That is a structural fact about the work, not a verdict on your company, and planning as though it were temporary is what turns it into a crisis.
Watch the cycle that catches most operators. You recruit only when someone quits, which means you are short the day you start looking. Being short means overtime and a dispatcher covering routes. Being short also means taking whoever applies, because you need someone on Monday. The person you settle for is more likely to leave, and you are back at the start, now with less money and a tired team.
A standing pipeline breaks that cycle, and it is less work than it sounds. A posting that never comes down. Applications reviewed on the same day every week, whether or not you have a seat open. And a warm bench: the people you interviewed and liked but had nowhere to put, kept in touch with rather than forgotten.
One number decides whether the pipeline works, and it is how long an applicant waits to hear back. Good drivers are applying to several companies in the same week and they take the first credible offer, so if your callback waits until Monday they have usually taken another job. Measure days to callback the way you measure days to payment, because they are the same kind of number: the delay is yours, and it is costing you something you can count.
The cadence
| What happens | How often | Who owns it |
|---|---|---|
| Applications reviewed and callbacks made | Weekly, even when fully staffed | Whoever runs the office |
| One backup works a full day in the seat they cover | Monthly | Owner |
| Separations logged by tenure band, with the exit question asked | Monthly | Owner |
| Depth chart updated: which seats are one deep | Quarterly | Owner |
| Role documents refreshed by the person in the seat | Quarterly | Seat holder |
Turnover has a shape
Turnover as a single percentage tells you nothing about what to fix. What helps is how long each person stayed before they went.
Inside 30 days
A hiring or onboarding problem
The job was sold as something it is not, or nobody trained them properly in week one. The labor market did not do this to you.
Three to six months
A schedule or pay problem
They learned the routes and then learned what the hours and the money actually are. This band tells you whether you were straight with them about the hours and the money.
After a year
Something changed
Long-tenure drivers leave when something concrete moved: a route, a dispatcher, a contract, a shift pattern. Find the change rather than blaming the person.
So log separations by band rather than counting them. Three people leaving in their first month and three leaving after two years are the same number and completely different problems, and only one of them is about the job market.
Then ask the departing person one question, and keep it to one: what would have kept you here? Five minutes, no form, and take the answer at face value even when it is uncomfortable. Turnover is one of the seven numbers on your scoreboard precisely because it moves before the others do. This is the section that tells you what to do when it moves.
Write the role down
The cheapest insurance available to an NEMT company is one page per seat. Count how many of your seats have one.
This is not a job description for a hiring ad. It is an operating document: what this seat does daily, weekly and monthly, what it decides on its own authority, what it escalates and to whom, and the handful of things that would confuse a new person in their first week. Logins, the facility that insists on a phone call, the broker portal that locks after three attempts.
Have the person in the seat write it, not the owner. Owners write what they believe happens, which is a tidier and less accurate document than the one written by the person who actually does it on a Thursday afternoon.
Then test it the same way you test the depth chart. Hand it to the backup on their practice day and watch where they get stuck. Every question they have to ask is a line the document is missing, and an hour spent that day is worth more than any policy binder.