The Cost of a Bad Truck Driver Hire: How to Calculate Yours
ON THIS PAGE8 sections
- Why does an industry average lose the budget meeting?
- What are the five cost lines you actually need to model?
- How do you build the model, step by step?
- What does the finished model look like on a real fleet?
- Where does the screening line belong in this model?
- What does this model not tell you, and where can it mislead?
- How do you check the model is actually right?
- Frequently asked questions
The owner asks what a screening budget buys. You say it prevents bad hires. He asks what a bad hire costs. You quote a number you read somewhere, he asks where it came from, and the meeting is over. The academic figure most often quoted on this — $8,234 per driver replacement, from the Upper Great Plains Transportation Institute at North Dakota State University — comes from carrier data for calendar year 1998. It is a real study with a real methodology, and it is also older than most of the trucks in your yard. The cost of a bad truck driver hire is not a number you look up. It is a number you build, from five lines you already have data for.
Why does an industry average lose the budget meeting?
An industry average loses because it is an average of carriers that are not you. The UGPTI study, The Costs of Truckload Driver Turnover, reported an all-company average of $8,234 per driver across the 15 carriers that supplied usable data, $7,923 for company-driver fleets, $8,612 for dry van company-driver fleets, and $6,420 for reefer company-driver fleets. Individual carriers landed anywhere from $2,243 to $20,729. The study is built on 1998 operating data, so every figure in it is a 1998 dollar attached to 1998 revenue per truck and 1998 insurance pricing. The authors also flagged their own limits: a small participant count, and most data items supplied by the companies themselves and therefore subjective and inconsistent.
The reasons two carriers land a factor of nine apart are structural, not mysterious. A fleet with spare tractors absorbs a departure differently than a fleet running every unit seated. A regional carrier hauling four short runs a day loses a different amount per idle day than a long-haul operation on a dedicated lane. A carrier with a $2,500 deductible carries a different tail risk than one at $25,000.
None of that makes the cost unknowable. It means the number has to be assembled from your own inputs. Where you have a real figure, use it. Where you don’t, use a range and label it as one — a defensible range beats a precise number nobody can source.
What are the five cost lines you actually need to model?
Five lines cover the overwhelming majority of what a failed hire costs, and each maps to a data source you already have. Build the spreadsheet with these rows before you touch a single number.
| # | Cost line | Formula | Where the input comes from |
|---|---|---|---|
| 1 | Empty seat | Idle days × (revenue per truck per day − variable cost per truck per day) | Dispatch/TMS revenue reports; fixed vs. variable cost split from your P&L |
| 2 | Recruiting and onboarding | Total recruiting spend for the period ÷ hires made in the period | Ad invoices, recruiter payroll, orientation pay, DOT physical and drug test invoices, MVR and PSP fees |
| 3 | Productivity ramp | Ramp days × (veteran miles/day − new-hire miles/day) × margin per mile | ELD or TMS miles by driver tenure |
| 4 | Safety and insurance tail | (Deductible × expected incidents attributable to the hire) + annualized premium change + workers’ comp exposure | Loss runs, broker renewal history, claim files |
| 5 | Compliance rework | Hours of admin per departure and re-hire × loaded hourly cost of the person doing it | Your own file audit; §391.51 requires the qualification file be kept for employment plus three years |
Two omissions are deliberate. There is no “lost customer” line, because unless you can name a specific account that left and say why, that number is a guess and it will be challenged. And there is no headline crash cost. National crash-cost averages are built for a different purpose than your P&L — they are population figures assembled for policy and regulatory analysis — so dropping one into a per-hire model overstates the case badly. Line 4 handles crash exposure through your deductible and your premium: your actual out-of-pocket.
How do you build the model, step by step?
Work the lines in order. Each step below names the action, whether it is required for a defensible number or merely smart, and the mistake that most often breaks the calculation.
Fix the window and count only the departures you mean
What you do: pick a 12-month window and pull every driver separation in it. Then split the list into three buckets — departures you replaced, departures caused by fleet growth or contraction, and departures where the driver’s own record or conduct was the reason (terminations for cause, no-shows, abandonments, drivers who failed a query or a check after hire).
Status: required. The whole model divides by this count.
The gotcha: the UGPTI methodology explicitly removes drivers hired for expansion from the replacement count, because those are growth costs, not turnover costs. If you leave growth hires in the denominator, your cost per replacement comes out artificially low and the model understates itself. Count replacements, not hires.
The third bucket — bad hires specifically, not general churn — is the one your owner cares about. General turnover includes the driver who retired and the driver who moved states. A bad hire is a seat you filled with someone you would not fill again. Track that subset separately; it is the population your screening budget is aimed at.
Price the empty seat
What you do: measure the days between the departing driver’s last dispatch and the replacement’s first revenue mile. Multiply by contribution per truck per day, not by revenue per truck per day.
Status: required, and usually the largest single line.
The gotcha: people multiply idle days by gross revenue and produce a number the CFO throws out in ten seconds, because a parked truck also burns no fuel and pays no driver. The correct multiplier is contribution: revenue per day minus the costs that stop when the truck stops. ATRI’s Analysis of the Operational Costs of Trucking: 2026 Update put the industry-average cost to operate a truck at $2.336 per mile in 2025, of which $1.854 per mile was non-fuel. Do not read either figure as the cost that keeps accruing while the seat sits empty. Driver pay and fuel are the two big lines that stop with the wheels, and tires and running maintenance stop with them. What keeps accruing is mostly the tractor and trailer payment, insurance, and permits and licenses — and that remainder is the only part that belongs in this line. Price the empty seat on contribution, not on total cost per mile.
If a spare tractor covered the freight, the empty-seat line is not zero but it is much smaller — it is whatever you gave up to cover it, usually overtime, a purchased-transportation move, or a service failure credit. Model that rather than pretending it didn’t happen.
Price recruiting and onboarding as a pooled cost
What you do: total everything you spent putting drivers in seats over the window, then divide by hires. Include job board and ad spend, recruiter salary and burden for the share of their time on driver hiring, referral bonuses, sign-on bonuses that vested, orientation pay and lodging, DOT physicals, pre-employment drug tests, MVR fees for every state the applicant held a license in, PSP pulls, and Clearinghouse query fees.
Status: required.
The gotcha: cost per hire is a pooled number, and the pool includes applicants who never got hired. You paid for the ad that produced eleven applicants and one hire. All eleven applicants’ screening costs belong in the numerator; only the one hire belongs in the denominator. Price it per-hire-only and you understate recruiting cost materially — often by more than the hired applicant’s own screening cost, since the ten who were screened and not hired came out of the same budget.
The regulatory floor is worth naming line by line, because these are real invoices. Under 49 CFR 391.23 you must obtain the driver’s motor vehicle record from every state licensing authority where they held a license in the preceding three years, and investigate their safety performance history with DOT-regulated employers for the preceding three years — both within 30 days of the date employment begins. Under 49 CFR 382.701 you must run a pre-employment Clearinghouse query before the driver performs a safety-sensitive function, query at least once a year for every covered employee, and retain each query and its results for three years — a retention burden the rule treats as satisfied automatically for an employer who maintains a valid Clearinghouse registration.
That Clearinghouse query is also how the drug-and-alcohol history requirement gets met. Since January 6, 2023, an FMCSA-regulated carrier satisfies the prior DOT drug and alcohol testing history requirement of 49 CFR 40.25 through the Clearinghouse as to other FMCSA-regulated employers. The separate manual §40.25 request — covering the two years before the application — still applies to previous employers regulated by a DOT mode other than FMCSA. You must obtain and review that information before the employee first performs safety-sensitive functions where feasible, and as soon as possible if it is not; and in no case may you let them keep performing those functions more than 30 days after their first safety-sensitive duty without having obtained the information or documented a good-faith effort. Every one of those steps has a per-applicant price. Our background screening walkthrough for trucking companies covers the operational sequence, and the DAC report guide explains what previous employers do and don’t put on the record you’re buying.
Price the productivity ramp
What you do: pull average miles per day for drivers in their first 30, 60, and 90 days, compare against your tenured-driver baseline, and multiply the gap by margin per mile for the length of the ramp.
Status: smart, not strictly required — but it is the line that separates a model from a guess.
The gotcha: the ramp is not only miles. A new driver runs fewer miles, but also generates more service failures, more re-consignments, more calls into dispatch, and more minor damage. UGPTI treated this as “production loss due to new drivers” — the revenue difference between what a new driver runs and what a veteran runs — plus a separate equipment maintenance line for the extra wear and fender-benders that come with drivers new to the equipment. Model miles only and you miss the rest of the ramp: the service failures, re-consignments, dispatch time and minor damage that come with a driver new to the equipment.
For a bad hire, this line behaves differently than for a good one. A good hire’s ramp cost is an investment that pays back over their tenure. A bad hire’s is a total write-off, because they leave before payback. That is the entire economic difference between turnover and a bad hire, and it is why the ramp deserves to be counted at full value in a bad-hire model and amortized in a general turnover model.
Price the safety and insurance tail
What you do: work from your loss runs. Take your deductible, multiply by the incident rate you actually observe among drivers in their first year, and add the annualized effect of any premium change your broker attributes to loss experience. Add workers’ comp exposure separately — UGPTI folded higher workers’ comp premiums for new drivers into its safety, insurance and legal cost category, alongside accident and legal costs.
Status: required if you want the number to survive scrutiny, because this is the line the owner intuitively believes is large.
The gotcha: do not reach for a national average crash cost. Those figures are assembled to value crashes at a population level for policy analysis, and using one as a per-hire input inflates the model to the point where nobody believes any of it. Your deductible plus your premium delta is a smaller number and a bulletproof one.
There is a second, slower tail: a driver’s violations attach to your CSA profile while they are on your authority. And serious traffic violations carry escalating disqualification consequences. Under 49 CFR 383.51, a second conviction for a serious traffic violation arising from a separate incident within three years — committed while operating a CMV, or in a non-CMV if the conviction results in revocation, cancellation or suspension of the driver’s license or driving privileges — disqualifies the driver for 60 days, and a third for 120 days. Major offenses run far longer: generally one year for a first conviction while operating a CMV, three years if the driver was operating a CMV transporting hazardous materials at the time, and a lifetime disqualification for a second conviction — though a State may reinstate after 10 years on completion of an approved rehabilitation program. A few offenses, including drug-trafficking and human-trafficking felonies committed with the vehicle, are lifetime on the first conviction with no reinstatement. A driver who arrives already carrying one serious violation conviction is one bad afternoon away from being unable to drive for you at all — and you are back to holding the empty seat.
Price the compliance rework
What you do: time-study your own paperwork. Count the hours it takes to close out a departing driver’s file and open a new one, and multiply by the loaded hourly cost of whoever does it.
Status: smart. It is usually the smallest line, but it is the easiest to defend because it is entirely internal.
The gotcha: the work does not end at the exit interview. Under 49 CFR 391.51 the qualification file has to be retained for as long as the driver is employed and for three years afterward, which means a bad hire generates a records obligation that outlives their employment by three years. And under 49 CFR 391.23, when your former driver applies elsewhere, you become the previous employer who has 30 days to respond to the new carrier’s safety performance history request — and if you have no safety performance data on that driver, you are still required to send a response confirming that none exists. Every failed hire adds a small permanent tax to your back office. A new-hire vetting checklist is the cheapest way to keep that tax from compounding.
Roll it up and pressure-test the total
What you do: sum lines 1 through 5 for a single representative bad hire, then multiply by your annual count of bad hires from Step 1 to get an annual exposure figure.
Status: required.
The gotcha: present both numbers, and present the per-hire figure as a range rather than a point. “Between $9,000 and $16,000 per failed hire, driven mostly by how long the seat sits empty” is a sentence that survives cross-examination. “$12,400 per failed hire” invites someone to find the one input they disagree with and dismiss the whole model.
What does the finished model look like on a real fleet?
Here is the shape of the output for a hypothetical 40-truck dry van fleet. Every input below is illustrative. They are placeholders showing how the arithmetic connects, not benchmarks — the entire point of the exercise is that you substitute your own.
| Line | Illustrative input | Illustrative cost |
|---|---|---|
| 1. Empty seat | 11 idle days × $420 contribution/day | $4,620 |
| 2. Recruiting and onboarding (pooled) | $38,000 annual spend ÷ 19 hires | $2,000 |
| 3. Productivity ramp | 45 days × 90 miles/day gap × $0.28 margin/mile | $1,134 |
| 4. Safety and insurance tail | $2,500 deductible × 0.35 expected incidents + $900 premium effect | $1,775 |
| 5. Compliance rework | 9 hours × $38 loaded hourly | $342 |
| Per failed hire | $9,871 | |
| Annual, at 8 failed hires | $78,968 |
Notice which line dominates. In this shape the empty seat is 47% of the total, driven by one variable you can measure precisely: days to fill. If yours is 11 and a better pipeline could get it to 6, that is worth more than any other single change in the model. Run your own numbers and the dominant line might be different — a carrier with a deep applicant pool and a high deductible may well find line 4 outweighs line 1. Either way, the model tells you what to fix, and gives you a denominator for every prevention dollar you want to spend.
Where does the screening line belong in this model?
The screening line belongs in the model as a sixth row, priced per applicant, and compared against one instance of the number you just built. That is the whole argument, and it does not require anyone to believe a vendor’s claim.
Add a row: cost per applicant screened × applicants per hire. If you screen eleven applicants to make one hire and the additional check costs a few dollars per applicant, that row is two orders of magnitude below the per-failed-hire figure in the table above. You are not claiming any particular prevention rate. You are showing that the ratio is lopsided enough that a modest prevention rate clears the bar, and letting the reader do the division themselves.
Now the honest part about what that row does and doesn’t buy. Your DOT-required checks — MVR, safety performance history, Clearinghouse, drug and alcohol history — are legally mandatory, and this model assumes you already run all of them. Nothing here is an argument for trimming any of it. What they are is a record of adjudicated events: convictions, positives, reportable accidents, formal terminations. They are excellent at catching a driver who has already been caught.
Those required checks are structurally blind to the pattern behind most of the failed hires in your Step 1 bucket. The driver who takes orientation pay and disappears, who no-shows on Monday of week two, who is unreachable on a Friday load, who leaves the tractor in a lot in Tulsa — none of that generates a conviction, a positive test, or a DOT-reportable event. It generates frustrated dispatchers at three previous carriers, and there is nowhere in the federally required stack for that to surface. The previous-employer investigation under §391.23 is a partial answer, but it is a 30-day-turnaround request — by letter, phone or interview — that many carriers answer with dates of employment only, even though the rule requires a substantive response or an affirmative confirmation that no safety performance data exists.
That gap is what driver reputation data from other carriers is for: a behavioral layer on top of the required checks — what dispatchers and safety managers at prior employers actually experienced, in a form you can read in the fifteen minutes you have. It is an addition to your compliance stack, never a substitute for any part of it. If the peer review layer and the formal check disagree, the formal check governs the compliance decision; the peer signal governs whether you want to spend eleven applicants’ worth of pipeline on this one. Price the compliance handling of that row honestly too — whatever you bring into a hiring decision carries handling obligations, and the limits below are the ones to check it against.
Priced honestly, that sixth row rounds to nothing against line 1. That is the comparison to put in front of the owner — not a payback period, just two numbers side by side. If the pattern you are pricing is specifically the driver who leaves the truck somewhere, our guide on cutting no-shows and abandoned trucks covers the operational side of the same problem.
What does this model not tell you, and where can it mislead?
The model is a budgeting tool, not a prediction. Four limits are worth stating out loud before you present it, because someone in the room will find them if you don’t.
It cannot attribute causation. The model tells you what a failed hire costs. It does not tell you that better screening would have prevented any specific one. Some failed hires are failures of dispatch, pay, home time, or equipment, and no amount of front-end vetting touches those. If exit interviews say drivers are leaving over home time, screening is not your highest-return fix and the model should not be used to argue otherwise.
Cost per hire is sensitive to the denominator. A fleet making six hires a year will see per-hire figures swing wildly on one unusual event. If your annual replacement count is in single digits, present a three-year rolling average.
Screening spend has a legal ceiling on what it can decide. When you use a consumer report for an employment decision, the FCRA applies. Per the FTC’s guidance for employers, before you get the report you must notify the applicant in writing in a stand-alone document — not inside the employment application — get their written permission, and certify compliance to the reporting company. Before an adverse action based on the report, you must give the applicant a notice including a copy of the report you relied on plus a copy of A Summary of Your Rights Under the Fair Credit Reporting Act. After the adverse action, you must give notice including the reporting company’s name, address and phone number, a statement that the reporting company did not make the decision, and notice of the right to dispute the information and get an additional free report within 60 days. The FTC also flags trucking-specific FCRA subsections at 15 U.S.C. § 1681b(b). Build those hours into line 5.
Do not assume the obligations stop at the report you think of as “the background check.” The rules turn on how information is compiled and what you use it for, not on the label a vendor puts on it, and some categories carry duties beyond the adverse-action sequence above. Before you bring any third-party information about an applicant into a decision — from any source, ours included — ask the provider which category their product falls into under the FCRA, and confirm the handling with your own counsel. That is a question for a lawyer who knows your process, not for a guide.
Nothing here is legal advice. Regulations are amended. Verify the current text of any section before relying on it, and run adverse-action procedures past your own counsel.
How do you check the model is actually right?
Validate it against money that already left the building. A model that cannot be reconciled to your P&L is a slide, not a tool.
Three checks, in order of strength:
- Reconcile line 2 to your general ledger. Total recruiting spend in the model should tie to the sum of advertising, recruiter payroll allocation, orientation pay, and screening invoices for the same period. If it doesn’t tie within a few percent, you missed a cost category or double-counted one.
- Reconcile line 1 to utilization. Total idle-truck days from departures should show up in your fleet utilization report as a gap. If the model says you lost 88 truck-days and utilization shows no such gap, you either covered the freight (re-price it per Step 2) or your idle-day count is wrong.
- Back-test one departure end to end. Take a single bad hire from last year, rebuild all five lines from the actual documents, and compare against what the averaged model predicts. Within about 20% and the model is good enough to budget against. Three times apart and one of your averages is hiding a long tail — usually idle days.
Rerun it annually, or after any change to your insurance program, pay package, or fleet size. Keep the assumptions visible in the spreadsheet next to the numbers they produce. The model wins the meeting not because the total is large, but because every input has a name and a source, and the owner can change any one of them and watch the answer move.
Frequently asked questions
How do you calculate cost per hire for truck drivers?
Total every dollar spent putting drivers in seats over a fixed period — advertising, recruiter payroll and burden, referral and sign-on bonuses that vested, orientation pay and lodging, DOT physicals, drug tests, MVR fees, PSP and Clearinghouse queries — then divide by the number of drivers actually hired in that period. Include screening costs for applicants who were never hired; the pool paid for them.
How do you calculate a driver turnover rate?
Divide the number of driver departures during a period by the average driver headcount over that same period, then multiply by 100. Use average headcount rather than a start-of-year or end-of-year snapshot, and exclude positions added or cut for fleet growth or contraction, or a growing fleet will show an artificially low rate.
What does an idle truck actually cost per day?
Contribution, not revenue: your revenue per truck per day minus the costs that stop when the truck stops — chiefly driver pay and fuel, plus tires and running maintenance. For scale on the total, ATRI’s 2026 Operational Costs report put the industry-average cost to operate a truck at $2.336 per mile in 2025 and $1.854 per mile excluding fuel. Only the fixed remainder — tractor and trailer payments, insurance, permits — keeps accruing while the unit sits.
Should crash costs go into a per-hire model?
Use your own deductible and premium history, not a national average crash cost. National crash-cost figures are assembled at a population level for policy and regulatory analysis, which is a different question than what a crash takes off your P&L. Multiplying your deductible by the incident rate you actually observe among first-year drivers gives a smaller number that will survive review.
Does a bad hire affect your CSA profile?
Yes — violations and inspections recorded while a driver is operating under your authority attach to your carrier profile, not the driver’s next employer. That is a lagging cost the empty-seat line does not capture. It is also why a driver’s existing record matters: under 49 CFR 383.51, a second conviction for a serious traffic violation from a separate incident within three years disqualifies a CDL holder for 60 days, and a third for 120 days. It applies to violations committed in a CMV, or in a non-CMV where the conviction costs the driver their license.
How long do you have to complete the required checks on a new driver?
Under 49 CFR 391.23, the motor vehicle record inquiry to each state where the driver held a license in the preceding three years, and the safety performance history investigation covering the preceding three years, must both be done within 30 days of the date employment begins. The prior drug-and-alcohol history under 49 CFR 40.25 runs on a different clock: obtain and review it before the driver first performs safety-sensitive functions where feasible, as soon as possible if not, and in no case may the driver keep performing them more than 30 days after that first safety-sensitive duty without it.
How do you justify a screening budget to an owner who won’t approve it?
Show two numbers side by side: your own per-failed-hire cost from the five-line model, and the annual screening spend expressed as cost per applicant times applicants per hire. Do not promise a prevention rate or a payback period. Let the ratio and the owner’s own judgment about how many bad hires it would need to catch do the work.
Is a per-hire cost model the same as a turnover cost model?
No, and mixing them is the most common error. A turnover model amortizes onboarding investment across a driver’s tenure, because a good hire eventually pays it back. A bad-hire model writes that investment off in full, because the driver left before payback. Run the same five lines both ways and keep the outputs in separate columns.