No numbers in this article. Not because turnover cannot be measured, but because we do not yet have our own data to publish and we are not going to borrow someone else's statistic and present it as if it described your depot. Everything below is structural: where the cost sits, and what moves it. When we have aggregate data worth publishing, it will appear here with that label on it.
Decompose the cost before arguing about it
"Driver turnover is expensive" is agreed by everyone and acted on by almost nobody, because the cost is never on one line of anything.
It arrives in four places.
Finding the replacement. Advertising, or the agency margin, plus the manager hours spent reading applications and interviewing people who do not start.
Getting them cleared. Right to Work, the licence check, whatever screening your policy requires, and the documents that arrive in pieces. This is elapsed time as much as money — days during which you are paying to cover the route another way.
Getting them useful. A driver who does not know the round is slower on it. That is not a criticism of the driver; it is the shape of the work. The gap closes over weeks, and during those weeks it shows up in the numbers.
The performance dip while all that happens. The uncovered routes, the newer driver's error rate, the manager attention diverted from everything else.
Add those and the fourth is usually the largest, and it is the only one that never appears on an invoice — which is exactly why turnover gets discussed as a recruitment problem rather than an operations problem.
The reason on the exit form is rarely the reason
People leaving a job they dislike are polite about it. Pay is the socially acceptable answer, it is unarguable, and it ends the conversation — which is what someone with one foot out of the door wants.
The reasons that do not get said out loud tend to be about predictability and standing.
Not knowing Thursday what you are doing Saturday, when you have a second job or a child to collect. Being told about a change by another driver rather than by the depot. Being marked down for something nobody explained. Asking a question and not getting an answer, twice.
None of those cost money to fix. All of them are about whether the operation behaves like it knows what it is doing.
The first two weeks decide it
A driver forms their view of an employer almost immediately, and almost entirely from process rather than from people. Everyone is friendly in week one. What differs is whether things work.
Consider two depots.
In the first, the driver is sent one link, uploads their documents once, is told what happens next and when, and is given a start date that does not move. The rota for their first fortnight exists. Someone shows them how to do the delivery photo on day one.
In the second, the driver is asked for a document, then a second one four days later, then the same one again because the first message was lost. The start date slips twice. The rota arrives the evening before. Nobody mentions the photo, and they find out it mattered when it is raised as a problem in week three.
Same pay. Same routes. Same vans. One of those drivers is still there in six months.
The second depot is not badly run. It is a depot where onboarding lives in someone's inbox — and inboxes lose things while the person who owns them is covering a route.
Predictable scheduling is a retention feature
Rota predictability gets treated as an operational nicety. For a workforce where a large share of people have another commitment — another job, childcare, study — it is the single most consequential thing you control.
A rota published late is not merely inconvenient. It converts a job that fits around someone's life into a job that does not, and the moment a driver has to choose between the two, you are not in the running.
Two concrete changes, neither of which costs anything:
Publish earlier and change less. A rota confirmed on Wednesday for the following week is worth more to a driver than the same rota confirmed on Saturday, even if the shifts are identical.
Tell the people affected, not everyone. A driver who receives five messages a week about changes that do not concern them stops reading messages — and then misses the one that does.
Our own driver scheduling page covers the mechanics. The retention point is simpler than the mechanics: the rota is the main way your depot communicates with people who are rarely in it.
Language is not a soft factor
A meaningful share of delivery drivers in the UK are not reading English as a first language. In practice that means a driver can be entirely competent and still be operating on partial information all day.
The consequences are not mainly about comfort. Half-understood instructions become mistakes, mistakes become performance conversations, and performance conversations with someone who did not fully follow the instruction in the first place feel like unfairness. That is how a good driver becomes a leaver without anyone naming a reason.
Communicating in the driver's own language, on a channel they already use, removes an entire category of avoidable friction. It is also the cheapest thing on this list.
Feedback that does not read as punishment
Every driver in a DSP is measured, continuously, by a system they did not choose. The question is not whether to give feedback — the numbers exist either way — but whether the driver hears about their work only when it is bad.
Three things separate feedback that improves performance from feedback that produces leavers.
Timing. A conversation about last week is a conversation about something nobody can change. Closer to the event is both fairer and more useful.
Specificity. "Your metrics are down" is a verdict. "These three stops on Tuesday" is a conversation, and it invites the explanation that might turn out to be a van problem or a route problem rather than a driver problem.
Proportion. If the only individual attention a driver receives is corrective, the rational response is to keep their head down — which is precisely the behaviour that stops you hearing about problems early.
Turnover shows up on the scorecard before it shows up in your headcount
This is the connection most depots make too late.
Almost every line on the Amazon scorecard is produced by a driver doing or not doing something at a doorstep. Photograph taken properly. Parcel not marked delivered when it was not. Customer not annoyed. Van driven calmly.
Drivers who have been with you a year do those things without being asked. Drivers in their third week do some of them, some of the time, depending on what they were shown on day one.
So a depot with high turnover is permanently running a rota with a higher proportion of new people on it, and the scorecard reads that as a performance problem. The briefing that follows is about delivery standards. The cause was that four people left in six weeks.
Our article on the Amazon DSP scorecard goes through what the published sources say each metric measures — including where they contradict each other. The retention point is the one underneath all of them: the metrics are a measure of experience on the round as much as of effort.
What to measure if you have no data
You do not need a system to start. You need three numbers, and you can reconstruct all three from records you already hold.
Days from accepted offer to first route. This is the one that surprises people. Measure it for the last ten drivers who started. If the spread is wide, the variance is your process, not your applicants.
Survival at thirty days and at ninety. Of the drivers who started in a given month, how many were still there a month later, and three months later? Two numbers, tracked monthly, tell you more than any exit interview.
Where the leavers were in their tenure. Everyone leaving inside the first month is a different problem from everyone leaving at month eight. The first is onboarding. The second is the job itself.
None of that requires software. It requires someone to write down the dates as they happen — which is the same discipline everything else on this page depends on.
What this adds up to
Retention in this industry is not won with pay rises, and the operators who grow are rarely the ones paying the most. It is won by being an employer that is predictable: documents collected once, start dates that hold, rotas published early, messages that arrive in a language the person reads, and feedback that is specific and timely.
Every one of those is an admin problem wearing an HR costume. Which is the encouraging part — admin problems are the kind you can actually fix.
When we have aggregate data on onboarding time and driver tenure across customers, we will publish it here and label it as ours. We would rather publish nothing than publish a number we cannot stand behind.
How WaveGo helps with this
Driver recruitment and onboarding that runs itself — five seats free, no card and no contract.