Tools · People analytics
Utility Analysis
The dollar value of a better hire or a real program effect — computed, not vibed.
The method
Utility analysis (Brogden–Cronbach–Gleser; Cascio & Boudreau's decision-science extension)
Every year, organizations spend real money on hiring processes and development programs whose value nobody can state in dollars. The budget conversation runs on conviction — HR believes the structured interview is better; Finance sees only the assessment invoice. The question the method answers is the one the CFO actually asks: what is a better hire worth, and does this program pay for itself?
The industrial psychologists who built utility analysis — Brogden, Cronbach, and Gleser, carried into practice by Cascio and Boudreau — made a claim that still surprises people: the dollar value of a selection procedure is computable from four quantities you can actually estimate. How valid the procedure is (its correlation with job performance), how selective you can afford to be, how much performance varies in dollar terms (one standard deviation of performance, SDy — for many jobs, roughly 40% of salary), and how long hires stay. Multiply them through and you get ΔU: the net dollar gain over hiring at random, minus what the assessment cost you.
Cascio and Boudreau's discipline in Investing in People is not the formula — it's the posture. Treat HR spending as investment under uncertainty, not cost to minimize. Their decision-science frame (and its strategic extension in Beyond HR) insists you also ask where better selection matters most: in pivotal roles, where a standard deviation of performance moves the business, not just where hiring volume is high. The method's honest critics note that SDy estimation is contested and validity coefficients travel imperfectly across contexts — which is why the serious practitioner reports a sensitivity band, not a point estimate.
In the book, this is the point where you'd be handed the formulas and sent off to build the spreadsheet. Here the calculator is live — the math runs deterministically in code (never estimated by a language model), and the write-up tells you which of your assumptions to challenge before you take the number to Finance.
The books behind this tool
- Investing in People: Financial Impact of Human Resource Initiatives — Wayne F. Cascio & John W. Boudreau
- Beyond HR: The New Science of Human Capital — John W. Boudreau & Peter M. Ramstad
How it works
Deterministic Brogden–Cronbach–Gleser selection utility (validity × selection-ratio ordinate × SDy × tenure − costs) and Boudreau/Cascio program utility (effect size d), with SDy via the 40%-of-salary rule or caller-provided, break-even effect sizes, ROI multiples, and ±25% sensitivity bands — all in code (the LLM never does arithmetic). The LLM interprets only: exec narrative, the method's contested assumptions as they apply HERE, and which inputs to validate before betting on the number. Grounded in the people-analytics corpus.
You bring
{ mode: selection|program, selection?|program?: {...}, context?, cluster? }
You get
{ computation (ΔU · ROI · break-even · sensitivity), interpretation (narrative · caveats · assumptions_to_challenge), grounded_in, provenance }
Use it for
- →Business case for structured selection: ΔU of replacing unstructured interviews, with the break-even validity
- →Training-program go/no-go: the effect size it must achieve to pay for itself
- →Pairs with turnover-cost for the full talent-economics story
See it work
example outputSelection utility of replacing unstructured interviews with structured interview + work sample: 400 applicants → 40 enterprise AE hires, r=.31, $350/applicant, 3-year tenure.
ΔU (net utility): $2,340,844.42 over 3 years · ROI 17.72× · break-even validity r = 0.0175
| SDy (40% rule) | $38,000 |
| Selection ratio | 0.1 |
| Mean z of hires | 1.755 |
| Total assessment cost | $140,000 |
Replacing unstructured interviews with a structured interview plus work sample for enterprise AE hiring shows a projected first-cohort net utility of $2,340,844.42 (ΔU per person-year of $19,507.04) against $140,000 in total cost — a 17.72x ROI. The dominant driver is the very large standardized effect of 1.755, which reflects the well-documented incremental validity gain (0.31) of structured methods over unstructured interviews combined with a highly selective 10% selection ratio. Even the conservative ends of the sensitivity band — $1,720,633.32 at SDy of $28,500 or incremental validity of 0.23 — remain strongly positive, and the break-even effect of only 0.0175 is trivially low relative to the 1.755 modeled here, meaning the decision would need almost the entire effect to evaporate before it stops paying off. For enterprise AE roles, where individual performance variance in bookings is genuinely large, this is a directionally credible bet, but the headline dollar figure should be read as an order-of-magnitude 'this clears the bar comfortably' signal rather than a precise forecast, given how much it rides on the SDy estimate and on actually realizing that 10% selectivity.
Caveats
- SDy was set via the 40%-of-salary rule of thumb ($38,000), the most contested input here — the sensitivity swing from $28,500 to $47,500 moves ΔU from $1.72M to $2.96M, and for enterprise AEs whose performance is heavily quota/territory-driven, real SDy could be far above 40% of salary, making even this large number potentially conservative — or the rule could overstate it if pay compresses true output variance.
- The 0.31 incremental validity comes from selection literature averages that travel poorly to a specific enterprise-AE context; local realities (territory quality, ramp time, deal cycles measured in quarters) may not let a work sample predict on-the-job selling as cleanly as it predicts in validation studies.
- The utility assumes top-down selection at a 10% selection ratio — this is an upper bound that only holds if your best-scoring candidates actually accept offers; in a competitive AE talent market, declined offers force you down the score distribution and erode the modeled gain.
- The model applies the per-person-year gain across an assumed tenure without accounting for effect decay — AE performance advantage from better selection may compress as tenure, coaching, and territory assignment converge, so multi-year totals are softer than the arithmetic implies.
Validate before betting on this number
- SDy = $38,000 (40% of salary): validate against actual AE performance-value dispersion (e.g., spread in booked revenue/margin across your current AE population) before trusting the dollar magnitude — this single input roughly determines whether ΔU is $1.72M or $2.96M.
- The 10% selection ratio: confirm you truly have a 10:1 qualified-applicant-to-hire funnel AND that top-ranked candidates accept — if offer acceptance is selective, the effective selection ratio worsens and the gain shrinks.
- Incremental validity of 0.31 over unstructured interviews: pressure-test whether your current 'unstructured' baseline is really as invalid as assumed and whether the new work sample is administered consistently — the low band of 0.23 still yields $1.72M, so this is less fragile than SDy but worth local validation.
- The number of hires and tenure horizon over which the per-person-year $19,507.04 is applied — confirm cohort size and expected AE tenure, since the total is a straight multiplication that decay and turnover will erode.
Run it on your data
Call it on your own inputs — over the API, or hand it to your AI agent via MCP. Discovery is open; running it is metered.