Real cost of a bad engineering hire
Real cost of a bad engineering hire
A bad senior engineering hire costs $600,000 to $1.2 million over 18 months. That figure includes severance, recruiter fees, vacancy during the search, code debt that compounds, peer disengagement, and the months you knew it was not working but had not acted yet. SHRM's 0.5x to 2x salary rule of thumb is the floor, not the ceiling.
Most cost-of-bad-hire articles stop at the spreadsheet line items: recruiter fees, signing bonus, severance. Those are the easy numbers. The damaging costs live in the gap between "we know this is not working" and "we did something about it." That gap is the real story.
The headline number: what the studies actually say
The most-cited benchmarks for cost of a bad hire come from three sources:
- US Department of Labor: minimum 30% of the employee's first-year wages.
- SHRM: 0.5x to 2x annual salary to replace any employee. C-suite turnover runs 213% of salary.
- CareerBuilder survey: 74% of companies report having hired the wrong person at some point; the average reported cost was $14,900.
- Toggl Hire 2025 Report: direct costs $5,000 to $10,000, indirect costs ballooning to $30,000 to $150,000+ per bad hire.
- Tech-specific data: for software engineering and cybersecurity roles, total cost can hit 50% to 200% of annual salary.
These numbers are correct but lossy. SHRM's 2x ceiling assumes you act decisively, replace within a normal hiring cycle, and the bad hire did not own customer-facing infrastructure. For a senior engineer in a startup, none of those assumptions usually hold.
Direct costs: what hits the P&L immediately
These are the line items your finance team will actually find on the books.
| Cost line | Mid engineer ($120k) | Senior engineer ($200k) |
|---|---|---|
| External recruiter fee (20%) | $24,000 | $40,000 |
| Signing bonus (burned) | $5,000 | $20,000 |
| Severance (2-3 months) | $25,000 | $50,000 |
| Interview-loop time (20-40 hrs @ $86-$150) | $2,000-$6,000 | $4,000-$6,000 |
| Vacancy during search (60-90 days) | $30,000 | $50,000 |
| Replacement recruiter fee | $24,000 | $40,000 |
| Direct subtotal | ~$110,000 | ~$200,000 |
For a deeper look at what engineers actually cost loaded, the software engineer total compensation breakdown walks through base + bonus + equity + signing math for 2026.
These are the costs every CFO sees. They are also less than half the real total.
The hidden costs nobody puts on the spreadsheet
Code debt that compounds
A bad senior writes 6 to 12 months of production code before anyone agrees the hire is not working. That code does not disappear when they do. It sits in your repo, generating bugs, blocking refactors, and forcing every future feature to route around its assumptions. Realistic cleanup cost on 6 months of bad senior output: 2 engineers working 4 to 6 weeks. At loaded full-time cost, that is $40,000 to $90,000 of pure repair work.
Peer engineer disengagement
This is the cost most companies refuse to measure. A bad hire on a 6-person team has two effects: the good engineers either fix the bad work (their own velocity tanks 20% to 40%) or they stop reviewing it carefully (quality across the entire repo drops). Gallup's 2025 State of the Global Workplace found managers explain 70% of the variance in employee engagement. One bad hire that the manager visibly tolerates broadcasts that performance does not matter here. Your top engineer notices. They start interviewing in month four.
Manager time tax
CareerBuilder's survey of CFOs found managers spend roughly 17% of their week directly supervising underperformers. On a manager pulling $250,000 loaded, that is $42,000 per year. For 6 to 9 months until action, $21,000 to $32,000 of management time, which would otherwise have gone to the people who are working.
Customer impact for senior bad hires on customer-facing systems
If your bad senior owns billing, auth, or the primary API, the cost stops being theoretical. One P0 outage costs an enterprise customer's confidence. Two costs the renewal. A botched migration to Stripe or Supabase can cost weeks of revenue. We have seen single-incident churn costs of $200,000 to $600,000 traceable to one bad senior hire who shipped a database migration that did not survive contact with production traffic.
The inertia tax: months you knew and did nothing
This is the single largest hidden cost and almost nobody quantifies it.
The pattern: 3 to 6 months in, the manager privately knows the hire is not working. Then 2 to 4 more months pass before action. Why?
- PIPs require documentation, legal review, and a 60-day runway.
- Losing the headcount feels worse than keeping the underperformer.
- The signing bonus and recruiter fee feel like sunk cost (they are not, but they feel that way).
- Q3 is "the wrong time" to fire someone. So is Q4. So is January.
During those 6 to 9 inertia months, you pay full salary for 30% to 50% productivity. On a $200,000 base, that is $80,000 to $120,000 of straight inertia waste, before any of the second-order costs above. The longer you wait, the more code debt and peer disengagement accumulates.
The honest read: the SHRM 6-to-9-month detection rule is not a useful guideline. It is a description of how slow most organizations are to act. Every week you compress that window is real money saved.
Worked example: $200,000 senior bad hire over 18 months
Pulling it together. Loaded annual cost is $260,000 (1.3x base for benefits, equipment, taxes, tooling).
| Cost category | Conservative | Realistic | Worst case |
|---|---|---|---|
| Direct (recruiter, severance, vacancy, interview time) | $150,000 | $200,000 | $250,000 |
| Inertia tax (7 months at 60% productivity loss) | $80,000 | $109,000 | $140,000 |
| Code debt cleanup | $40,000 | $90,000 | $150,000 |
| Peer engineer drag (3 peers, -30% velocity, 9 months) | $120,000 | $176,000 | $230,000 |
| Customer-facing incidents and churn | $50,000 | $200,000 | $600,000 |
| Recruiter fee on replacement + ramp | $50,000 | $80,000 | $120,000 |
| 18-month total | ~$490,000 | ~$855,000 | ~$1.49M |
The realistic case for one bad $200k senior is $600,000 to $1.2 million, full stop. The worst case approaches $1.5 million when you include lost enterprise renewals.
Why better interviews cannot save you
The reflex response is to tighten the funnel. More technical screens, more behavioral panels, more take-homes. The data does not support this as a complete fix.
The Schmidt and Hunter (1998) meta-analysis of 85 years of personnel selection research is the standing reference. The validity ceiling for predicting on-the-job performance:
| Selection method | Validity coefficient (r) |
|---|---|
| Years of experience | .18 |
| Education level | .10 |
| Unstructured interview | .38 |
| GMA (general mental ability) alone | .51 |
| Work sample | .54 |
| Structured interview alone | .51 |
| GMA + structured interview | .63 |
| GMA + work sample | .63 |
| GMA + integrity test | .65 (ceiling) |
A validity of .65 means 35% of the variance in actual performance is not predicted by your hiring process. Translation: even with the best-evidence selection methodology in the world, you get a bad hire roughly 1 in 5 times. The 2021 updates to Schmidt-Hunter actually revised some of these estimates downward, suggesting the real ceiling is closer to .60.
You cannot interview your way to zero bad hires. The mathematical ceiling will not let you. What you can do is shorten the consequence window so that when a bad hire slips through (and one will), the damage stops at $1,500 instead of $1.2 million.
The how long does it take to hire a software engineer in 2026 piece walks through what each interview step actually buys you in predictive lift.
What changes when bad-hire risk approaches zero
Every cost in this article is downstream of one assumption: that ending the employment relationship is slow, expensive, and emotionally fraught. Change that assumption and the math collapses.
Cadence is on-demand engineering booking. Every engineer is AI-native by baseline, vetted on Cursor / Claude Code / Copilot fluency before they unlock bookings. You book by the week. Pricing is locked: Junior $500/week, Mid $1,000/week, Senior $1,500/week, Lead $2,000/week. If a senior is not working out, you do not write a PIP. You do not pay severance. You rate them down on the daily check-in, the system surfaces a replacement, and you have a new engineer onboarded by Monday. Median time to first commit on the replacement is 27 hours, drawn from a 12,800-engineer pool.
Worst-case downside on a bad Cadence booking: $1,500. One unused week of senior billing. Compared to the $600,000 to $1.2 million downside on a full-time bad senior, that is a four-to-five-orders-of-magnitude difference in risk.
If you want to model what a project actually costs with weekly billing versus full-time, run the numbers on the Cadence ROI calculator before your next hire. It accounts for ramp, severance probability, and replacement cost. We built it because the conventional cost-per-hire calculators do not.
Honest framing: this is not the right answer for every role. If you are hiring a Staff engineer to build institutional knowledge over 5 years, you want headcount. The on-demand model wins for project work, bridge capacity, anything where the scope is bounded by quarters not decades. For a deeper breakdown of when each model wins, see our engineering hiring market in 2026 deep dive.
Comparison: the bad-hire downside by hiring path
| Hiring path | Bad-hire cost (18 months) | Time to replace | Severance risk | Notes |
|---|---|---|---|---|
| Full-time senior ($200k) | $600,000 to $1,200,000 | 60-120 days | 2-3 months pay | SHRM baseline plus inertia tax |
| Toptal senior contractor | ~$25,000 to $50,000 | 2-4 weeks to re-match | Contract notice, often 2 weeks | Honest: better than FTE, slower than weekly |
| Cadence senior ($1,500/wk) | $1,500 maximum | 2 minutes to rebook, 27-hour median first commit | None | Risk floor is one unused week |
Toptal is genuinely a better answer than a full-time hire if your downside is what you are optimizing for. Their network is mature, their vetting is real, and the contract structure gives you more flexibility than W-2. Where we differ: their effective rates run higher than Cadence senior pricing, and their notice periods (typically two weeks) still build a small consequence window.
Decision framework: how to actually use this data
Before signing the next senior offer, walk these five questions:
- Is this a 12-week scope or a 5-year capability? Twelve-week scopes never deserve full-time hires.
- What is my real downside? Multiply base by 4 to 6 for senior roles, not 0.5 to 2. Then ask if you can absorb that.
- Have I shortened the consequence window? Three-month contractor agreements beat 18-month employment agreements on this dimension alone.
- Am I paying senior comp for mid scope? Most "senior" job descriptions describe mid-level work. The Stack Overflow developer survey 2026 highlights shows the median mid engineer with 2 years of experience is now shipping work that required senior comp in 2022.
- Do I have a realistic exit path before I sign the offer? If the answer is "we will figure it out," you are pre-paying the inertia tax.
FAQ
What is the average cost of a bad engineering hire?
SHRM puts the replacement cost at 0.5x to 2x annual salary. For a $200,000 senior software engineer, the realistic 18-month true cost is $600,000 to $1.2 million once you include code debt, peer disengagement, customer impact, and the months between recognition and action. Tech roles run higher than SHRM's general benchmark because the second-order code-debt costs compound.
How long does it take to spot a bad engineering hire?
On average, 3 to 6 months to recognize the problem clearly and another 2 to 4 months to act on it. That 6-to-9-month gap is the single largest hidden cost. Daily ratings and weekly retros (the model used on weekly-billed platforms) compress detection to days instead of quarters.
Are recruiter fees worth paying for senior engineers?
External recruiter fees run 15% to 25% of first-year base, or $30,000 to $60,000 on a $200,000 role. Worth it if your funnel is dry and you have a high-conviction job spec. Not worth it if you can route through a network, a referral pool, or a weekly-booking platform that has already done the vetting work.
Can better interviews prevent bad hires?
Only partially. The Schmidt-Hunter meta-analysis caps validity around r = .65 even for the best combination of cognitive testing, structured interview, and integrity testing. That means a perfect process still produces a bad hire roughly 1 in 5 times. Tightening the gate has diminishing returns; shortening the consequence window does not.
How does weekly billing change bad-hire economics?
It collapses the downside from $600,000-$1.2 million to roughly $1,500. On Cadence, an underperforming senior engineer can be replaced any Friday with no severance, no notice, and no PIP paperwork. The risk floor of any booking is one unused week of billing. That is the structural difference, not a marketing claim.