The ROI of Pre-Employment Assessment, Calculated Honestly
The ROI of a pre-employment assessment is a calculation, not a statistic — and the defensible version of it has been in the industrial-psychology literature for decades as the Brogden–Cronbach–Gleser utility model. In plain terms: the annual value a selection method adds equals the number of people you hire, times the validity of the method (r), times the rupee spread between a good and an average performer in that role (SDy), times how selective you can afford to be (Z), minus what the assessing costs you. Three of those four inputs you already know or can look up. The fourth — SDy — is the one nobody measures precisely, and it is the reason published ROI claims differ by an order of magnitude while sounding equally confident. The most-quoted figure in this category, that a bad hire costs 30% of first-year salary, is attributed almost universally to the US Department of Labor and, in our own search, to no traceable primary document; SHRM's own accounting puts the same cost at 50–200% of annual salary. When two authorities differ by nearly seven times, the number is not your evidence. Your own inputs are.
Selection utility
The money value a selection method adds compared with hiring at random from the same applicant pool. It rises with three things — how many people you hire, how selective you can be, and how much better a strong performer is than an average one in that specific role — and with the validity of the method, which is the correlation between what the assessment measures and later job performance. Utility is a per-role figure: the identical assessment has a large ROI for a role where performance varies widely and a negligible one where everybody performs about the same.
What the public materials actually say
Every line below was taken from the vendor's own site and is cited. Where something is absent, it is dated rather than asserted as a permanent limitation.
The validity numbers most vendors quote are out of date. A 2022 re-analysis found decades of personnel-selection meta-analyses had systematically over-corrected for range restriction, and revised the estimates downward: cognitive-ability tests from .51 to .31, work samples from .54 to .33, structured interviews from .51 to .42.
Source: Sackett, Zhang, Berry & Lievens (2022), Revisiting Meta-Analytic Estimates of Validity in Personnel Selection, Journal of Applied Psychology 107, 2040–2068
That revision re-ordered the field. Structured interviews are now the strongest single predictor of job performance in the set, ahead of cognitive-ability tests — the reverse of the ranking repeated in most assessment marketing.
Source: Sackett et al. (2022), as above
Validity is a correlation, not a hit rate. A validity of .42 does not mean 42% of your hires will succeed, and it does not mean the method is right 42% of the time. It means the method's ranking of candidates and their eventual performance move together to that degree — which is enough to shift the average quality of a cohort, not enough to guarantee any individual hire.
Source: Standard interpretation of a validity coefficient; see kaairo.ai/science
The "a bad hire costs 30% of first-year salary" figure is attributed to the US Department of Labor across dozens of vendor and agency pages, none of which cites a primary DOL publication. SHRM's fuller accounting of the same cost is 50–200% of annual salary. Both are repeated as authoritative; they differ by roughly seven times.
Source: Searched August 2026 — every result traced was a secondary citation, no primary source located
Assessment cost is the input people get wrong in the safe direction. It is not just the per-candidate fee: it includes the time your team spends reviewing results, and the candidates you lose to a longer process. A cheap assessment that adds a week to time-to-hire can cost more than an expensive one that removes three days.
Source: Definitional — the cost term C in the utility model above
Side by Side
The estimates in the left column are the ones still quoted in most assessment marketing. The right column is the 2022 correction. If a vendor's ROI case rests on the older figure, its arithmetic is inflated at the input.
| Older estimate (Schmidt & Hunter, 1998) | Revised estimate (Sackett et al., 2022) | What that means for your calculation | |
|---|---|---|---|
| Structured interview | .51 | .42 — now the strongest single predictor in the set | The smallest downgrade of the three. A structured, rubric-scored interview holds up best under the correction. |
| Cognitive ability test | .51 | .31 | The claim that general mental ability is the best predictor of job performance no longer holds on these numbers. |
| Work sample | .54 | .33 — the largest single drop | Still a strong method, but an ROI case built on .54 overstates the value added by roughly a third. |
Work out your own inputs
Three of the four inputs are yours, not ours. These calculate them from your numbers — no sign-up, and no figure of ours is baked in.
Estimates the cost of a mis-hire from your own salary, tenure and replacement figures, rather than applying a borrowed percentage. This is the honest way to approach SDy: not to look it up, but to bound it with numbers you can defend.
Best for: Putting a defensible rupee figure on the downside you are trying to avoid.
Source: kaairo.ai/tools/cost-of-a-bad-hire-calculator
Works out what an unfilled role costs per day. This matters to assessment ROI in both directions: a screening step that removes days from your process adds value, and one that adds days subtracts it.
Best for: Sizing the speed half of the trade, which utility models usually ignore.
Source: kaairo.ai/tools/cost-of-vacancy-calculator
Works out how many candidates you need to see for a given confidence of finding a strong one. This is the selectivity term (Z) in the formula — the input that moves ROI most and the one teams most often leave to chance.
Best for: Setting the selection ratio deliberately instead of by whoever applied.
Source: kaairo.ai/tools/how-many-candidates-to-interview
Running the calculation, in four steps
- 1
1. Fix the role, not the company
Utility is per-role. Run it for one role with real volume — the one you hire twenty of, not the one you hire one of. An assessment's ROI on a role you fill twice a year is close to noise regardless of how good the assessment is.
- 2
2. Bound SDy rather than looking it up
The money spread between a strong and an average performer is the input nobody can measure exactly. A common rough proxy is 40% of annual salary, but treat that as a bound to test, not a fact: run the calculation at a pessimistic and an optimistic value and see whether your decision actually changes. If it does not, you have your answer without needing the precise number.
- 3
3. Use a current validity figure
Use the 2022 estimates, not the 1998 ones (table above). If you are comparing methods, note that the correction changed the ranking, not just the magnitudes — a structured, rubric-scored interview now outranks a cognitive test on these numbers.
- 4
4. Count the full cost, including days
Add the per-candidate fee, the reviewing time, and the effect on time-to-hire priced with your cost-of-vacancy figure. An assessment that improves selection but adds a week to the process can be net negative in a competitive market — and that is a real finding, not a reason to skip the step.
Frequently Asked Questions
What is the ROI of pre-employment testing?
There is no single number, and any page that gives you one without asking about your role, your hiring volume and your performance spread is quoting someone else's inputs. The value added is the number you hire × the validity of the method × the rupee spread between a strong and an average performer × how selective you can be, minus the cost of assessing. The two variables that dominate the answer are hiring volume and performance spread — which is why the same assessment can be transformative for a 200-person frontline intake and pointless for two senior hires a year.
How do I measure the ROI of pre-employment assessments?
Measure it forward and backward. Forward: run the utility calculation above for one high-volume role, with a pessimistic and an optimistic SDy, before you buy. Backward, once you have data: split your existing hires by whether they cleared the assessment bar and compare their actual performance ratings, retention at six months, and ramp time. The backward version is the one that convinces a finance team, and it needs nothing more than the assessment scores you already hold and the performance data HR already keeps.
Is the '30% of salary' bad-hire figure reliable?
Treat it as a rule of thumb. It is attributed almost everywhere to the US Department of Labor, but we could not trace a primary DOL publication behind it, and SHRM's own accounting of the same cost is 50–200% of annual salary. A figure whose credible range spans seven times is not a basis for a business case. Use your own salary, tenure and replacement numbers instead.
Does a higher validity score mean better ROI?
Only in proportion, and only if the other inputs hold. Validity enters the formula linearly, so moving from .31 to .42 raises the value added by about a third — meaningful, but smaller than the effect of doubling your hiring volume or of being able to select the top 20% rather than the top 60%. Selectivity and volume move ROI more than method choice does, which is why the cheapest ROI gain is usually a wider top-of-funnel rather than a better test.
How long before an assessment pays for itself?
For a high-volume role it is usually the first cohort, because the cost is per-candidate while the value is per-hire and persists for as long as the person stays. For low-volume senior hiring it may never pay back on utility alone — and that is a legitimate result. Senior assessment is often bought for defensibility and consistency of process rather than for measurable ROI, and it is more honest to say so than to construct a business case out of borrowed statistics.
Where does Kaairo sit in this?
Kaairo runs structured, rubric-scored assessments — case studies, situational judgement tests and real-time AI voice interviews — which is the method family that held up best in the 2022 correction. Pricing is published in INR at kaairo.ai/pricing, so the cost term in your calculation is a number you can read rather than one you have to request. We deliberately do not publish an ROI multiple for our own product: we would be choosing your SDy for you, and that is the input this whole page argues you should not accept from a vendor.
Measure before you move someone
Define the target role's competencies and have the people you are deciding about actually demonstrate them — with pricing you can read right now, no sales call required.
More Buyer's Guides
This buyer's guide is published by Kaairo. SHRM, and the platforms named and all other product names are trademarks of their respective owners; Kaairo is not affiliated with or endorsed by them. Competitor information is drawn from each vendor's public materials as of August 2026 and may have changed — please verify current details with each vendor before making purchasing decisions.