Three operational conversations are on your calendar this quarter, and none of them have good answers yet. Your CFO will ask why your 2026 plan needs twelve senior hires when tech is laying off everyone. Your team will ask whether to allow AI in technical interviews, and you don't have a defensible stance yet. And the staff engineer who just had coffee with an Anthropic recruiter will give notice within sixty days unless you have a retention move that isn't an equity refresh.

The headline number, since you'll need it for the budget conversation: hiring twelve Senior/Staff engineers at a 50-person scaleup costs roughly $3.6–3.9M in Year 1. Your recruiting budget shows about $400K of it. The other ~$3.4M is real and is spread across engineering interviewer time, onboarding ramp drag, and the cost of the offers you'll have to turn down. None of it lives on a single dashboard, which is why the conversation with your CFO has been going badly.

This paper is the sourced math for those three conversations. Built on government statistical agencies, academic research, primary cap-table and payroll records, and engineering-leadership statements from companies that have no commercial interest in saying their own hiring is broken. No vendor pitch. Source methodology and limitations at the bottom for anyone who wants to check the work.

1. The 2026 scaleup labor picture is split in two

The dominant 2025–2026 narrative is "tech labor surplus." It is wrong at the scaleup level in two specific ways: senior-engineer supply is tighter than the headlines suggest, and scaleups absorbed the majority of 2025 layoff pain.

The big picture is clear. BLS Information sector employment is down 342,000 jobs (-11.0%) from its November 2022 peak through April 2026 (BLS Employment Situation, April 2026). JOLTS Information openings in March 2026 were down 33% year-over-year (BLS JOLTS, March 2026). The Information-sector layoff and discharge rate doubled to 2.4% in March 2026 from 1.3% a year earlier, the sharpest acceleration of any private sector. CompTIA's tech-occupation unemployment rate climbed from 2.8% in June 2025 to 4.0% in November 2025 and ~3.5% in April 2026 (CompTIA Tech Jobs Report, monthly).

The supply data shows seniors are the tight part of the market. The Federal Reserve Bank of New York's labor market tracker shows CS graduate unemployment at 7.0% and computer engineering at 7.5% versus the all-recent-grads rate of ~5.7% (NY Fed Labor Market for Recent College Graduates, Q1 2026 update). At the same time, CS underemployment sits at 16.5%, far below the 41.5% all-grads underemployment baseline. The split: high unemployment among new CS graduates, but those who do land a CS job are properly employed. The entry-level part of the market is hard; the senior end is short of supply.

The bootcamp pipeline that historically supplied juniors collapsed in parallel. 2U/Trilogy shut down all university-partnered bootcamps in December 2024 (~96,000 cumulative graduates over the prior 5 years); Epicodus, Momentum Learning, Turing School, and SNHU all closed between 2023 and 2025 (Class Central; Inside Higher Ed). The aggregate junior-supply pipeline narrowed at the same time companies stopped hiring at the entry level. CRA Taulbee 2024 reported all-time-high US/Canada CS PhD production of 2,352 (+8.2% YoY) and +9.9% new-student-enrollment growth, but at the doctoral and undergraduate enrollment level, not at the fresh-bachelor's level (CRA Taulbee Survey 2024 Annual Report).

The second misread in the headlines is where the layoffs actually landed. Crunchbase News and TechStartups industry analysis of Layoffs.fyi data show roughly 60% of 2025's 122,549 tracked tech layoffs came from early- and mid-stage startups (Series B–D), not Big Tech (TechStartups, Q3 2025). The Big Tech cuts grab headlines (Meta announced ~10% / ~8,000 starting May 2026 per CNBC; Microsoft cut ~15,000 across 2025; Amazon ~30,000 corporate roles since October 2025 per Computerworld 2026 timeline). The actual employment damage is concentrated at the scaleup tier the VPE reading this is operating in.

Carta's primary cap-table data shows the scaleup retreat directly. Average headcount at Series D startups fell 29% from the 2023 peak to 131 employees by 2025. Series B average dropped from 53 to 45 employees; Series A from ~25 a few years ago to 16 (Carta State of Startup Compensation H1 2025). Series A is now a 16-person company on average. The "lean by default" norm has spread three rungs down the funding ladder. Carta's January 2026 net-hires data shows VC-backed startups made 26,030 new hires versus 20,378 departures (a 1.3× ratio), the slowest January since 2018. The comparable January 2022 ratio was 3.8×.

SignalFire's 2025 State of Talent report puts numbers on the new-grad collapse and the AI-talent concentration. New-grad share of hires at startups fell to below 6% (down from ~25% pre-pandemic); Big Tech sits at 7% (SignalFire State of Talent 2025; data drawn from primary LinkedIn and payroll records). 65% of US AI engineers are based in San Francisco + New York combined, and Anthropic, OpenAI, and Meta are growing engineering teams 2–3× faster than they're losing them. Engineers are 8× more likely to leave OpenAI for Anthropic than the reverse, and ~11:1 from DeepMind to Anthropic. The frontier labs are soaking up senior engineers fast enough to make the scaleup squeeze worse for any company hiring AI-adjacent backend, ML, or infrastructure engineers.

BLS still projects +15% growth for the combined software developer / QA / tester family 2024–2034 (BLS Occupational Outlook Handbook, current edition), but the projection hasn't yet captured the 2024–25 contraction or the AI productivity shock. Treat it as an upside scenario, not the baseline. What's actually happening at scaleups in this cycle is the leading signal, and that signal is demand piling up at the senior level inside an overall market that's shrinking.

2. What a senior engineer actually costs at a scaleup

Compensation is the part of the cost we can pin down with independent data. Carta's cap-table records and Levels.fyi's community-sourced compensation data are the closest non-vendor primary sources.

Carta H1 2025: average new-hire engineering salary across Carta-tracked startups is ~$189,000 (tied with product as the highest function) (Carta State of Startup Compensation H1 2025). The figure spans all engineering levels at all startup stages on Carta's platform. BLS Occupational Employment Statistics for Software Developers (15-1252) puts the May 2024 median annual wage at $133,080, with the top decile at $211,450 and the bottom decile at $79,850 (BLS OEWS, May 2024). The gap between the BLS national median and the Carta startup mean reflects the scaleup-tier premium: VC-backed startups pay above the national distribution because they hire above the level distribution.

Levels.fyi's 2025 End of Year Pay Report puts the US Senior SWE median total compensation at $312,000 (+4.2% YoY); the AI-engineer premium runs +14.2% (Senior) to +18.7% (Staff) over non-AI peers (Levels.fyi 2025; AI Engineer Compensation Trends Q3 2025). Senior+ compensation at named scaleups, all from Levels.fyi's community-sourced compensation pages (continuously updated 2025–2026):

These figures matter because they set the lowest competing offer your seniors will weigh against yours. The AI-lab ceiling ($563K Senior at Anthropic; $633K Staff at Stripe) is what the senior engineer at a Series B–D scaleup is comparing against when they take a recruiter call. Even if the typical scaleup pays $200–250K Senior, the candidate's reference point is the friend who took the AI lab offer at twice the package. Carta's Q2 2025 data confirms the AI premium at the startup level: median AI/ML engineer salary grew +5.4% to +9.1% from January 2024 to June 2025 depending on startup valuation band (Carta Q2 2025 AI engineer compensation).

Equity grants at scaleups have not recovered from the post-2022 reset. Carta data shows equity grants at startups remain ~26% below pre-2022 levels; product/engineering grants are ~15% smaller than the year prior (Carta H1 2025). The first-engineering-hire equity median sits at ~1.5% (4-year vest); this drops to ~0.85% at hire #2 and ~0.33% by hire #5 (Carta data via SaaStr). By Series A, Senior-eng grants typically run 0.3–0.8% of fully-diluted equity; by Series B, 0.2–0.7% (Index Ventures Rewarding Talent; Stock Option Counsel).

The cash-cost and the equity-cost both matter because the sum is what a senior engineer weighs against an outside offer. At a Series B–C scaleup paying $190K base + 0.4% equity on a $300M post-money valuation, the four-year package is roughly $760K base + ~$1.2M equity face value (heavily dependent on liquidity and exit timing) ≈ $2M nominal. Anthropic Senior at $563K/year × 4 = $2.25M with materially less liquidity friction. Equity-rich scaleups still win on upside but lose on certainty, and 2024–2025 trained engineers to discount certainty heavily.

The cost-per-hire components beyond compensation include several lines that the independent record can quantify, plus a few that remain dominated by recruiting-vendor benchmarks. What we can pin down:

• Engineering interviewer time: Ashby's recruiter-productivity benchmark (250K+ hires, ~14M apps across venture-backed scaleups under 1,500 employees, methodology disclosed) puts technical roles at 17.9 interviews per hire and 24.7 hours of internal team time per hire, the highest of any role family. Ashby's startup-specific cut shows 21–29 hours of interviewer time per technical hire at scaleups in the 25–300 employee band. At a $150/hour loaded engineering rate, that's ~$3,700–$4,400 of engineering OpEx per hire, spread across the team and never tied back to the recruiting budget. (Ashby Recruiter Productivity Trends Report.)
• Loaded compensation: Standard CFO accounting practice loads benefits, payroll taxes, tooling, and overhead at 25–40% above base salary, putting a $190K Senior at roughly $238–266K fully loaded.
• Onboarding ramp: Senior engineers take 3–9 months to ramp to productivity based on what experienced operators report (HackerNoon technical-leadership analysis; Pragmatic Engineer onboarding-cohort reporting); this lines up with the BLS Employee Tenure Survey's 4.9-year median for architecture-and-engineering occupations.
• Replacement cost on a wrong hire: The Center for American Progress's 2012 academic meta-analysis pegs replacement cost at 16–213% of annual salary, with the high end specifically applying to highly skilled long-term employees (Center for American Progress 2012). For a Senior engineer at $190K, that's a $30K–$405K cash-equivalent event when a wrong departure happens.
• Offer-acceptance gap: Ashby's 230,000-offer dataset puts the technical-role offer-acceptance rate at 73%, eleven points below the 84% business-role rate. Engineering candidates carry more offers at once and run the longest negotiations, which means a 12-senior plan needs ~16–17 offers to land 12 hires.

What we still can't pin down at any confidence: agency-fee distribution, recruiter productivity at the role/seniority level, channel-mix conversion rates, time-to-fill at the role/seniority level. These metrics are dominated by recruiting-vendor benchmarks beyond the two cited here.

A hired-and-onboarded Senior engineer at a Series B–C scaleup is a multi-hundred-thousand-dollar commitment in cash, opportunity cost, and embedded option value. A wrong hire that exits within 18 months is a high-six-figure mistake.

3. The interview signal collapsed

Remote technical interviewing, the mechanism that's supposed to manage hiring signal, broke between 2023 and 2026. The strongest single piece of evidence is a controlled experiment. The study put senior engineers (4+ years experience) through standard technical interviews with and without ChatGPT assistance, with the interviewers blinded to which group each candidate was in. Verbatim LeetCode questions: pass rate moved from 53% (control) to 73% (with ChatGPT). Custom questions not on LeetCode: pass rate dropped to 25% in the cheating condition. Zero of 32 interviewers detected the cheating in any case. (interviewing.io, "How hard is it to cheat with ChatGPT in technical interviews," n=32 completed interviews, control-vs-treatment design.) The methodology is unusually rigorous for the recruiting industry. Most data on how widespread cheating is comes from vendor surveys, not controlled experiments. The takeaway: the only reliable defense against AI cheating in remote technical screens is custom problem design that doesn't appear in training data, and most companies have not done it.

Anthropic's own engineering blog confirms the conclusion from the opposite direction. The Anthropic take-home was completed by ~1,000 candidates over ~18 months and produced "most of their current performance engineering team," but by May 2025 Claude 3.7 Sonnet outperformed more than 50% of human candidates on it within the 4-hour window. The take-home has been redesigned three times to stay ahead; the latest version uses constraint-based puzzles (Zachtronics-inspired) (Anthropic Engineering, "Designing AI-Resistant Technical Evaluations," 2025). The model that broke the test is built by the same company that admitted it.

The market response has split into two clean stances:

The middle path, officially banned with no detection, is the worst possible state and is where most scaleups actually sit. Both alternatives (clean-ban with in-person, or explicit-allow with redesigned format) are defensible. The middle is not.

The Roy Lee / Cluely scandal made the problem visible to non-technical executives. A Columbia undergraduate built a Chrome tool ("Interview Coder") that screenshots LeetCode problems and feeds them to ChatGPT, then used it to obtain interview success at Amazon, Meta, TikTok, and Capital One; Amazon rescinded his offer; Columbia suspended him for one year (NBC News, March 2025). He raised $5.3M seed (April 2025) and a $15M Series A from Andreessen Horowitz (June 2025), $20.3M total, under the explicit "Cheat on Everything" tagline (TechCrunch, April + June 2025). Lee later admitted his publicly-claimed $7M ARR was actually $5.2M, a 35% fabrication on top of the cheating product itself (TechCrunch, March 2026).

The North Korean angle is the real-world threat for any scaleup hiring remote. The number of US companies that hired North Korean IT operatives using AI-generated personas grew 220% in the past 12 months; 320+ companies have been infiltrated (Fortune, August 2025). Remote-only engineering hiring loops where deepfake video plus AI-driven coding answers pile on are the attack surface. Series B–D companies with distributed teams and lean security/HR are the highest-risk profile.

The academic foundation for what works is older but solid. Sackett et al. 2022 meta-analysis (re-evaluated by Wingate 2025) puts structured-interview real-world predictive accuracy at r = 0.42, the highest of any common hiring tool. Unstructured interviews predict job performance only barely better than a coin flip (Wingate et al., International Journal of Selection and Assessment 2025; Sackett et al. 2022). The takeaway is that the interview format most resistant to AI cheating (structured, custom-question, ideally in-person, scoring-rubric-driven) is also the format that produces the most reliable signal independent of cheating considerations. The two priorities point the same way.

Pragmatic Engineer's reporting captures the operator view from outside the vendor world. Engineering managers and directors at startups and mid-sized companies told Gergely Orosz that hiring is harder in 2025 than in the 2021 hot market, primarily due to AI-generated candidate floods. Maestro.dev was cited as one example of a small company overwhelmed by AI-fabricated applicants (Pragmatic Engineer, "The Reality of Tech Interviews in 2025"). Stack Overflow Developer Survey 2025 confirms the candidate-side baseline: 84% of professional developers report adopting AI tools; 51% use them daily (Stack Overflow Developer Survey 2025, n=49,000+). The cheating problem is structural. AI tooling is now baseline for the candidate pool, and the interview format has not adapted.

4. Equity stopped working as a retention tool

The retention mechanism that historically held senior engineers in place, equity grants vesting over four years toward an expected payout at exit, is weaker than it used to be. Carta's primary cap-table records are the cleanest non-vendor data we have.

Stock option exercise rates by startup employees collapsed from 54.2% in Q4 2021 to 32.2% in Q4 2024 (Carta, "Should I Exercise My Vested Stock Options?" 2024). Fewer than one in three vested grants is being exercised. The signal: employees no longer believe in eventual liquidity at the strike price implied by the 409a, and an unexercised option is a much weaker handcuff than a vested-and-exercised position with capital at risk.

Refresh grants, the second mechanism, are inconsistently applied at scaleups. ~20% of startup employees received a refresh by year 1; ~50% by year 2; refresh grants average 30% of an equivalent new-hire grant (Carta Annual Equity Report 2024). Series B–E companies allocate 40–50% of their equity pool to refresh grants. That's material spend, but only at companies that have built the discipline to run it.

Tender offers have stepped up to fill the gap, and the volume is real now. Carta executed 396 tender offers in 2025, up 62% YoY, totaling $18.4 billion in US startup tenders (Carta Tender Offers Q2 2025). Tender participation rose from 36.6% (Q1 2021) to 56% (H1 2025); subscription rates rose from 73.8% to 99.9%. By stage: 69% participation at Series D, 78% at Series E+. When a tender is offered, employees take it 56% of the time; when the company opens the order book, it gets 99.9% subscribed.

The broader signal is that secondaries are now standard, not the exception. VC secondary transaction value reached $61.1 billion in the 12 months ending June 2025, more than the combined VC-backed IPO value of $58.8 billion over the same period (Carta VC Secondary Trends Q2 2025). For the first time, employees and early investors are getting more liquidity from secondaries than from IPOs. The tender has become the main thing keeping seniors from leaving, because the equity carrot isn't golden anymore, and the surrounding market built the infrastructure to make it work.

5. The retention math at scaleups

Overall engineering attrition looks healthy and hides two distinct groups. Median US tenure across all wage and salary workers fell to 3.9 years in January 2024 from 4.6 in 2014; architecture and engineering occupations specifically: 4.9 years median tenure (BLS Employee Tenure Survey, September 2024 release). Carta-tracked startups saw voluntary turnover drop 31% YoY by end of 2024; voluntary departures were down 40% from the June 2022 peak (Carta State of Startup Compensation H2 2024). The "Great Stay" is real and engineers are participating in it.

The overall number hides a long tail. The wrong engineer leaving costs 16–213% of annual salary, with the high end specifically applying to highly skilled long-term employees per the Center for American Progress 2012 meta-analysis of 11 academic papers over 15 years. At a scaleup paying $190K average for Senior engineers, one regretted Staff exit is plausibly a $190K–$405K cash-equivalent event under the academic distribution. And at a 35-IC engineering team, losing the wrong Staff IC also redistributes 200+ hours of context across the remaining team.

The structural trigger is well-known. Roughly 50% of startup employees leave within ~3 years, often clustered at the 4-year vesting cliff completion (Carta cohort data); 43.4% of startup employees hired in 2021 left within two years (Carta via Data Driven VC). Vesting completion remains a leading indicator of departure even with the equity-mechanic shift described above.

The push factors show up clearly in surveys. Stack Overflow Developer Survey 2024 (n=65,437): only 19% of professional developers are happy with their current job; 48% are "complacent." The 2025 update moved happy slightly to 24.5% with not-happy at 28.4%; 46% of developers are "not looking" for a new role, leaving 54% open to opportunities, with 14.8% "considering strongly" (Stack Overflow Developer Survey 2025). 22% of developers report critical burnout; only 21% are healthy (LeadDev survey, March 2025, n=617 engineering leaders, independent community-published).

The manager factor is the most concerning data point. Gallup's State of the Global Workplace 2025 reports global employee engagement fell to 21% in 2024 from 23% in 2023, the first annual fall since 2019–2020. Manager engagement collapsed from 30% (2023) → 27% (2024) → 22% (2025). 56% of managers are looking to leave their current job vs 50% of non-managers. Managers account for 70% of the variance in team engagement, and they are now the most likely cohort to walk. Disengaged employees cost the global economy $438 billion in lost productivity in 2024 per Gallup.

The boomerang trend is the leading indicator of how the senior-IC market is shifting. 35% of all new hires in March 2025 were former employees, up from 31% YoY and 27% in 2018 (ADP Research, "Boomerang Hiring Makes a Comeback," 2025; based on millions of US W-2 records). In the tech (information) sector specifically, 68% of new hires in March 2025 were boomerang, roughly double the prior year's 34%. Boomerang 3-year retention is 64% versus 45% for fresh hires; salary premium for returning employees is 25–28%. Scaleups that maintain alumni relationships are quietly winning the senior-retention war from outside the funnel.

The senior-engineer-specific retention problem is the main risk for scaleups. SignalFire reports Anthropic, OpenAI, and Meta are growing engineering teams 2–3× faster than they're losing them, while Tesla, Bloomberg, and Walmart are losing engineers faster than they hire (SignalFire State of Tech Talent 2025). The share of engineers with 4+ years at one company dipped only slightly from 2015 to 2024 (59% → 52%). Long-tenure cohorts are surprisingly resilient. The two-group split is real: most engineers either leave fast or stay long.

Reaching Senior Software Engineer typically takes 5–8 years at most tech companies (Will Larson, Staff Engineer: Leadership Beyond the Management Track, 2021). The IC-vs-management split happens at this level. The single most-cited reason senior engineers leave, according to Pragmatic Engineer, Will Larson, Camille Fournier, and Charity Majors, is "the only path forward is management." Companies with parallel IC tracks (Staff/Principal/Distinguished) materially reduce regretted attrition; SignalFire's tenure data shows engineering organizations with formalized IC tracks retain Senior+ at higher rates.

6. Worked scenario: a 12-senior hiring plan at a 50-engineer scaleup

Setup: Series C SaaS company. 50-person engineering org. ARR ~$60M. Plan: hire 12 Senior/Staff engineers in the next 12 months. Average $190K base for Senior hires (Carta H1 2025 anchor). Loaded fully-burdened IC cost ~$260K (37% loaded markup, standard CFO accounting practice). Hybrid + remote-allowed.

Cost build, anchored only on independent figures:

The cash commitment (and near-cash items) for 12 hires lands at roughly $3.6–3.9M in Year 1 (compensation + load + ramp + sign-on + interviewer time), before any agency fees or recruiter loaded cost. Equity dilution at 4.8% of post-money valuation is a separate strategic cost that Boards and CFOs typically track; the cash-equivalent value depends entirely on exit timing.

The funnel volume the plan needs to drive. With Ashby's tech offer-acceptance rate of 73% (230K-offer dataset), 12 accepted hires requires ~16–17 offers extended. Each offer typically requires 17.9 interviews per hire (Ashby technical benchmark). That's ~300 engineering-hours of interviewer time across 12 hires, more if the offer-decline rate worsens. What we still can't pin down at any confidence: agency fee distribution, channel-mix conversion at the senior level, and time-to-fill drift quarter-over-quarter. These metrics are dominated by ATS-vendor and search-firm benchmarks beyond the Ashby and interviewing.io data cited above.

The risk-adjusted view requires planning around regretted attrition. Carta's median annual turnover rate for equity-holding employees runs at 17.5% (Carta H1 2024). A 12-person new-hire cohort plausibly loses 1–2 in the first 24 months at a $95K–$285K replacement cost per exit per the Center for American Progress academic meta-analysis. The single-exit downside is a quarter-of-revenue event at the unit-economics level for a $60M-ARR company.

For FY26 OpEx planning, what this means in practice is that the cash side of a 12-senior hiring plan is comfortably north of $3.6M in Year 1 before adding agency fees and recruiter loaded cost. A plan that under-models the sourcing, interviewing, and ramping cost typically misses by 15–30% of the visible cash budget. The independent floor is the compensation, ramp, and Ashby-anchored interviewer-time cost; the additional vendor-specific overlay (recruiter productivity, agency mix, channel conversion) is what determines whether the plan lands at $3.9M or $4.5M+.

7. What this means for FY26

Seven moves a scaleup VPE can make this year that are supported by the independent record:

1. Plan compensation budgets against the AI-lab ceiling, not the BLS median. The relevant outside offer for a senior engineer at a Series B–D scaleup is not the BLS national-median $133K but the Anthropic Senior at $563K or the Stripe Staff at $633K (Levels.fyi 2025–2026). Even if the typical scaleup pays $200–250K Senior, what the candidate is comparing your offer against is the AI lab paying twice the typical scaleup package.

2. Treat tender offers as a primary retention lever, not an exception. Carta executed $18.4B in US startup tenders in 2025 with 99.9% subscription rates and 56% participation when offered (Carta Tender Offers Q2 2025). This is now infrastructure. A scaleup that runs a structured tender every 18–24 months from Series B onward holds senior engineers more reliably than equity refreshes alone, which face the falling-exercise-rate problem (Carta's Q4 2021 → Q4 2024 exercise rate fell from 54.2% to 32.2%).

3. Build a Staff/Principal IC track at Series B, not Series D. What experienced engineering leaders have written (Will Larson; Camille Fournier; Pragmatic Engineer; SignalFire tenure data) is consistent: companies with parallel IC tracks retain Senior+ at higher rates because management isn't the only career path forward. The cash cost is near-zero; the retention upside is multi-hundred-thousand-dollar per regretted exit avoided per the Center for American Progress 2012 academic meta-analysis.

4. Pick a clean stance on AI in interviews. Either ban AI in screens AND mandate an in-person final round (Cursor's 2-day work trial; Google + McKinsey's one-in-person policy; Stripe's live-coding-only loop), or explicitly allow AI AND redesign the format for it (Canva's mandate; Meta's E4–E5 pilot Oct 2025). The middle path is where most scaleups actually sit, and is what Anthropic's engineering blog admits is broken at the take-home level. Both alternatives are defensible; the middle is not.

5. Plan capacity for senior fills measured in months, not weeks. BLS Information sector employment loss, NY Fed CS-grad unemployment, and Carta startup-headcount contraction together signal that the 2026 senior-engineer market is tight in a lasting way. Plan against multi-month senior fills, not multi-week. The cost of getting it wrong is one quarter of revenue per missed seat at scaleup engineering productivity.

6. Run a base case AND a slip case in every quarterly board update. The independent floor (compensation, ramp, replacement cost) already gets to multi-million-dollar Year-1 commitment for a 12-senior plan. The extra real-world layer that includes recruiter time, agency fees, and time-to-fill drift can plausibly add 15–30% on top. Showing the board the slip case proactively in Q1 is materially cheaper than explaining the overrun reactively in Q3.

7. Treat alumni and boomerang re-hires as a primary channel. ADP's primary W-2 records show 68% of new tech hires in March 2025 were former employees; boomerang 3-year retention is 64% vs 45% for fresh hires (ADP Research 2025). This is a net-positive channel that costs near-zero to maintain (alumni network, low-friction rehire process, explicit invitation back at tender events), and that scaleups underinvest in relative to its measured yield.