We found no sourced figure for the cost of a bad hire that applies to a software engineer. Two widely repeated numbers are not what their sources say, and the one survey that does carry a real dollar figure covers all roles and is nine years old. So the useful move is to build your own number from your salary, your timeline and your team's hours.
What the quoted figures actually trace to
Start with the $17,000 figure that some pages, INOP among them, attribute to CareerBuilder. CareerBuilder's own December 2017 release says $14,900. It came from a Harris Poll run between 16 August and 15 September 2017, with 2,257 hiring managers and HR staff and 3,697 full-time workers; the release gives no role breakdown. The same release put the cost of losing a good hire at $29,600 and found that 74% of employers had made a bad hire. Useful as a signal, but it's dated and says nothing about engineers.
The other famous figure is that a bad hire costs 30% of first-year salary, attributed to the US Department of Labor. Business.com and HBK both state it and neither links a document. INOP's sources section says no current Department of Labor publication states it and the original has never been publicly traced, though the same page attributes it to the DoL elsewhere. Our own searches found no Department of Labor document either. The figure is widely repeated, which is different from being sourced. Treat it as a rule of thumb, not data.
Then there's the range of $17,000 to $240,000. Veremark attributes it to Business News Daily, so that's a second-hand number and we haven't traced it further. HBK adds a replacement-cost multiple of salary (one-half to two times) attributed to SHRM, with no link. We've left those out of the model below.
What you can safely say is narrow: one 2017 survey of all roles put the average at $14,900, and we couldn't find an engineering-specific number that traces to a primary source.
Why engineering is different
Generic figures miss the line items that make an engineering mistake expensive. This section is judgment, not measurement, so we're not putting numbers on it.
A bad engineer consumes other engineers. Reviewers spend more time on each pull request, seniors pair on work that should have been routine, and the manager runs extra check-ins and writes the improvement plan. That time comes out of your best people, whose hours are the most expensive on the team.
Code is also a lasting cost. A weak hire's output stays in the codebase after they leave, and someone has to find it, rework it or live with it. Add access and security cleanup when they go, and the delay to whatever they were supposed to ship. A late feature that blocks a sales commitment costs more than the salary behind it.
The extreme case is the person who never had the skills they claimed. If you suspect that, read what to do if the hire turns out to be fake.
How to calculate the cost of a bad engineering hire
Every input below is yours to supply. The example numbers are made up round figures for illustration, not benchmarks, and shouldn't be quoted as averages for engineers.
1. Fix the loaded cost of the role
Take annual salary plus employer costs (social contributions, equipment, tooling, benefits) and divide by 12. This is your monthly loaded cost. For illustration, say €120,000 a year loaded, so €10,000 a month.
2. Count the months until you act
Count from the start date to the day you made the decision to part ways, not the day the contract ended. Include onboarding. For illustration, 4 months. Then decide what share of that pay bought work you didn't need. We'll assume half, so 4 x €10,000 x 50% = €20,000.
3. Add the replacement cost
Use a real quote: a recruiter fee, a buyout or the internal cost of running a new search. For illustration, a 20% fee on €100,000 base salary, so €20,000. If you hire through a partner that replaces an engineer at no extra recruitment cost, this line drops, but check what's covered in writing.
4. Add manager and reviewer hours
Ask the manager and the two engineers who reviewed most of the work for a rough tally of extra hours. For illustration, 100 hours at an internal rate of €80, so €8,000.
5. Add the delay cost
Estimate how long the roadmap slipped and what a month of delay costs you in revenue, a missed deadline or a delayed funding story. This is the hardest input and it can be the largest. For illustration, a two-month slip at €7,500 a month, so €15,000.
6. Sum it and compare with prevention
Add the lines: €20,000 + €20,000 + €8,000 + €15,000 = €63,000 for this made-up case. Now price a prevention step. For illustration, a two-week paid trial at the same loaded rate costs about €4,600; we round to €5,000. Divide: 5,000 / 63,000 is a bit under 8%. In this example, the trial pays for itself if it prevents one bad hire in every twelve to thirteen hires it touches (63,000 / 5,000 = 12.6). Your numbers will differ, and the break-even is the point of the exercise.
Where bad hires come from
CareerBuilder's 2017 survey also asked employers why their bad hires happened, and many of the answers point to vetting or process gaps. 33% said the candidate lied about their qualifications. 35% said the person lacked the skills but seemed quick to learn. And for 45% of employers, a bad hire meant the worker's skills didn't match what they'd claimed when hired. 30% felt pressure to fill the seat quickly, 25% ignored warning signs, and 10% had not run a complete background check.
Read that list as a process audit. In our judgment, lying and skill mismatch are what a work-sample test or a live technical session catches. Rushing and ignored warning signs are process failures that a fixed, written set of stages makes harder to skip, because nobody gets to drop a step because the seat has been empty for three months. The survey is old and covers every industry, so use it for the pattern rather than the percentages. The same release reported that employers saw lower productivity (37%), time spent recruiting and training a replacement (32%) and compromised quality of work (31%).
How vetting and a paid trial reduce the cost
The expected cost of a hire is the cost of a failure times the odds of it. Vetting works on the odds, and the cost model above tells you how much to spend.
Verify identity and experience before you make an offer. Test the skills with work that resembles the job, not a puzzle. Our guide to vetting software developers covers the full sequence, and designing a technical screening process goes deeper on the early stages.
A paid trial is, in our judgment, the strongest single test, because it measures what the interviews only predict: how the person works in your repo with your reviewers. It also caps the downside. If the trial fails, you've spent two weeks instead of four months. Some vetted marketplaces offer trial periods of their own, and there's a comparison of how trials and vetting differ between Lemon.io and Arc.dev if you're weighing that route.
HighCircl runs four vetting stages led by senior engineers (background and experience verification, communication and product thinking, a take-home project mirroring real work, and a live architecture session), and about 1 in 10 applicants passes. That's one example of a structured process, not proof the numbers above apply to anyone else.
If a hire is already failing, the clock matters more than the model. There's a separate walkthrough on deciding on a failing senior hire before probation ends.
What this means for your hiring process budget
Your decision is how much to spend per hire on vetting and a trial, against accepting the risk that some hires fail. Don't pick that number from a blog post, including this one. Run the six steps for your last bad hire, using real payroll and calendar data, and you'll have a figure your CFO can check.
Compare the price of a trial or extra assessment stage with your calculated failure cost. The CareerBuilder reasons (false claims, skill mismatch, rushed decisions) are the failure modes such a step is meant to address. If you're buying through a vendor instead of hiring directly, ask what happens when the engineer doesn't work out, and get the replacement terms in writing, since that clause changes line three of your model. For how to judge a vendor on those terms, see how to evaluate a nearshore vendor in seven steps.
Frequently asked questions
What counts as a bad hire?
A bad hire is someone whose real cost to the team, in pay, other people's time and delayed work, exceeds what they delivered, to the point that you end the relationship or wish you could. CareerBuilder's survey left the definition to the employer. Set your own threshold before you calculate, for example "left or was let go within 12 months", so the numbers are comparable between cases.
Is the "30% of salary" figure from the Department of Labor?
It's widely attributed to the Department of Labor, but we found no Department of Labor publication that states it. INOP's sources section says no current publication states it and the original has never been publicly traced, though the same page attributes it to the DoL elsewhere. Treat it as a rule of thumb, not data, and use your own calculation instead.
How much does a bad software engineer hire cost?
There's no sourced benchmark for engineers. The only dollar figure we could verify is CareerBuilder's $14,900 per bad hire, from a 2017 survey of all roles. For a specific number, follow the six steps above with your own salary, timeline and delay cost.
Does a paid trial reduce the cost of a bad hire?
In our judgment, yes. A trial cuts the months you pay for the wrong person from several to two weeks, and it tests the skills claims behind the 33% of bad hires where the candidate lied about qualifications, per CareerBuilder. Compare its price with your calculated failure cost to see the break-even.