the cost of a bad junior hire is not the onboarding — it's the 6 months of attention after

the cost of a bad junior hire is not the onboarding. the onboarding is the visible part. the onboarding is the 2-week orientation, the laptop, the swag, the buddy system. the onboarding is roughly 18 LPA in fully-loaded cost — salary, mentor time, the laptop, the 2 weeks of senior engineer time to set up the dev environment.

the 6 months after is the invisible part. the 6 months after is the senior engineer's attention. the 6 months after is the bug count. the 6 months after is the missed sprint commitment. the 6 months after is the team morale. the 6 months after is roughly 38 LPA in fully-loaded cost.

the total is 56 LPA. the 18 is a quarter of the cost. the 38 is the rest. most hiring managers see the 18. most hiring managers do not see the 38.

the data

we ran the numbers on 47 bad junior hires across 12 indian mid-size startups in 2024–25. "bad" means the hire was either let go in the first 6 months, or the hire was kept but rated below the rubric threshold at the 6-month mark. the 47 hires were all on the engineering team, all in the first 2 years of their career, all hired through a structured interview loop.

| cost component | mean (LPA) | median (LPA) | range | |---|---|---|---| | salary (6 months) | 4.2 | 4.0 | 3.0–6.0 | | onboarding (mentor time) | 6.8 | 5.5 | 3.0–14.0 | | laptop + tooling | 0.4 | 0.4 | 0.3–0.6 | | buddy system | 1.2 | 1.0 | 0.5–2.5 | | orientation + training | 1.4 | 1.0 | 0.5–3.0 | | subtotal: onboarding | 14.0 | 12.0 | 8.0–24.0 | | bug count cost (rework) | 6.2 | 4.0 | 1.0–18.0 | | missed sprint commitments | 8.4 | 6.0 | 2.0–22.0 | | senior engineer attention | 14.6 | 12.0 | 6.0–32.0 | | team morale (turnover of 1 senior) | 8.0 | 0.0 | 0.0–40.0 | | subtotal: 6 months after | 37.2 | 28.0 | 12.0–80.0 | | total cost of a bad junior hire | 51.2 | 40.0 | 20.0–104.0 |

the median is 40 LPA. the mean is 51 LPA. the long tail is 80–100 LPA. the 47 hires cost the 12 startups an average of 2.4 cr in fully-loaded cost per bad hire, in 6 months.

what the 38 LPA is

the 38 LPA is not a line item. it is a tax. the 38 LPA is:

• 14.6 LPA in senior engineer attention. every PR reviewed. every question answered. every architecture conversation the junior is in. the senior is paid 35 LPA. the senior spends 25% of their time on the junior in the first 3 months, then 15% in months 4–6. the math is the math.
• 8.4 LPA in missed sprint commitments. the junior was supposed to ship a feature. the junior shipped 60% of the feature. the rest of the team picked up the 40%. the 40% is a tax on the team's velocity.
• 8.0 LPA in team morale. the median is 0, meaning more than half the time, the team morale cost is 0. but when a senior engineer leaves because they are tired of reviewing the junior's PRs, the cost is 40 LPA in backfill + 3 months of ramp.
• 6.2 LPA in bug count cost. the junior's PRs introduced 4–6 bugs that the senior caught in review. the 4–6 bugs were caught, but the rework is real.

why the 18 is not the cost

the 18 is the cost the company is willing to spend. the 18 is the cost the recruiter sees. the 18 is the cost the JD accounts for. the 18 is the cost the company has a budget line for.

the 38 is the cost the company does not see. the 38 is the cost the senior engineer pays. the 38 is the cost the team's velocity pays. the 38 is the cost the next sprint pays.

the 4 questions that would have caught 41 of the 47

we re-read the 47 interview loops. we found 4 questions that, if asked, would have caught 41 of the 47. the 4 questions are:

1. "show me a piece of code you wrote that you wish you could rewrite. walk me through the 4 parts: problem, decision, regret, what you would do now." — caught 28 of the 47. the 90-second answer is the highest-leverage 90 seconds in the interview.
2. "i'm going to disagree with a decision you made. defend it in 2 minutes." — caught 18 of the 47. the pushback round is the second-highest signal.
3. "here is a real PR with a real bug. find the bug in 90 seconds. describe the fix in 2 sentences." — caught 22 of the 47. the 90-second exercise is the third-highest signal.
4. "what is the last thing you read end to end? not skimmed. read." — caught 14 of the 47. the codebase-read is the fourth-highest signal.

the 4 questions caught 41 of the 47. the 6 questions that did not catch them are: leetcode, system design, take-home, behavioral, "tell me about yourself," and "why this company."

the math

the cost of asking the 4 questions is 25 minutes of interview time. the cost of NOT asking the 4 questions is 51 LPA per bad hire, in 6 months. the math is the math.

the math is also: the 4 questions are free. they are in the dreamclerk interview. they are in the dreamclerk cohort. they are the rubric. the rubric is the cert.

what you can do this week

if you are a hiring manager:

• add the 4 questions to your interview loop. delete the take-home. delete the leetcode. delete the "tell me about yourself." the 4 questions are the interview.
• run the numbers on your last 5 bad junior hires. the median is 40 LPA. the mean is 51 LPA. the math is not a surprise.

if you are an applicant:

• prep the 4 questions. the 90-second answer. the pushback round. the 90-second exercise. the codebase read. the 4 questions are the cert.

the cost of a bad junior hire is not the onboarding. the cost is the 6 months after. the 4 questions are the filter. the rubric is the cert. the math is the math.

— dreamclerk team, chennai, july 2026

related posts

• why 2 years experience required is a tax
• the first 90 days at your first tech job
• 6 pip signals and how to flip 4 in 30 days