What adverse impact means for a small business adding a hiring assessment

Federal law starts holding you to the same nondiscrimination standard as a Fortune 500 company at 15 employees. Several states get there sooner, some as low as three or five. If you run a front desk, a couple of coordinators, and a small crew, you’ve probably already crossed that line without anyone telling you.

That matters the moment you add a hiring assessment. A personality test, a situational judgment test, or any structured screening step is a “selection procedure” under federal law, the same way a resume screen or an interview is. And selection procedures can produce adverse impact: a pattern where one group clears your process at a meaningfully lower rate than another, whether or not anyone involved meant for that to happen.

Here’s the thesis. Adverse impact is real, it can apply to you at a size you’d never expect, and the advice you’ll find online (calculate your four-fifths ratio, run a significance test) mostly assumes you have the candidate volume to make that math mean anything. You probably don’t. What actually lowers your risk at small-business scale isn’t passing a hidden statistical test. It’s building a process you could explain to a stranger if they asked.

What adverse impact actually means

Adverse impact isn’t the same thing as intentional discrimination. Nobody has to mean it for it to count. It just means a selection step, resume screen, interview, or assessment, produces a real difference in pass rates between groups defined by race, sex, or another protected characteristic.

The government’s definition comes from the Uniform Guidelines on Employee Selection Procedures, a 1978 regulation that’s still the baseline for how the EEOC evaluates hiring tests today. It doesn’t ask what you meant. It asks what your process actually did.

That’s why “the assessment doesn’t ask about race or age” isn’t the reassurance most owners assume it is. A test can be entirely neutral on its face and still produce adverse impact if one group passes it at a meaningfully lower rate than another. Outcome is the test, not intent. It’s the same reason AI resume screening can produce a skewed shortlist without anyone writing a discriminatory rule into it.

Why you can’t assume you’re too small for this

Title VII, the federal law behind most of this, covers employers with 15 or more employees for at least 20 calendar weeks in the current year or the year before. Below that headcount, the EEOC has no authority to reach you directly. Above it, you’re covered the same way a company ten times your size is.

The federal line comes faster than it feels like it should

Fifteen employees sounds like a lot until you count a front desk, two coordinators, a sales team, and a warehouse lead. Seasonal swings make it worse. Cross 15 for 20 weeks in the current year or the prior one, and you’re covered for that whole period, even if you dip back below it later.

State law usually sets the bar lower

Several states extend their own fair employment laws to much smaller employers than federal law does. California’s Fair Employment and Housing Act reaches employers with five or more employees. Connecticut’s state law reaches employers with three or more. The specific number varies by state, so this isn’t a substitute for checking your own, but the pattern is consistent: state law rarely waits for you to hit 15.

Put those two things together and “we’re small enough that this doesn’t apply to us” is a bet most owners in the 10 to 50 employee range are already losing, whether they know it or not. (This is general information, not legal advice. Talk to an employment attorney about what applies in your specific state and situation. If any of your hiring touches candidates in the EU, the rules layer differently again, and our breakdown of the EU AI Act and hiring covers that separately.)

The four-fifths rule works differently at your scale

The rule most adverse impact content leans on is the “four-fifths rule,” and it’s worth knowing in plain terms because a lot of vendor content treats it like the whole answer.

The rule of thumb, and its real limits

Under the Uniform Guidelines, a selection rate for any race, sex, or ethnic group that’s less than four-fifths (80%) of the rate for the highest-scoring group will generally be treated by federal enforcement agencies as evidence of adverse impact. Say 60% of male candidates pass your assessment and only 40% of female candidates do. That’s a ratio of about 0.67, below the 0.80 line, and it would generally read as a flag.

But the guidelines are explicit that this is a rule of thumb, not a legal conclusion. Smaller differences can still count as adverse impact if they’re statistically and practically significant. Larger differences might not count at all if they’re based on small numbers that could plausibly have happened by chance.

The small-numbers problem nobody mentions

Here’s the part that changes everything at your scale. The federal guidance itself gives a worked example: an employer selects three men and one woman out of a pool of 20 men and 10 women. Run the four-fifths math and it flags adverse impact. But the actual number of people selected is so small that the guidance says you can’t reasonably conclude adverse impact happened at all. It could just as easily be noise.

That’s not an edge case for a small business. That’s a typical hiring round: a couple dozen candidates, three or four offers. The ratio that “generally” signals adverse impact for a company processing 400 candidates a month can flip on one or two hiring decisions when your pool is this thin, in either direction. A clean ratio doesn’t prove you’re fine. A flagged one doesn’t prove you’re not. At small volume, the math is often just too noisy to trust either way.

The recordkeeping exemption isn’t a legal pass

Here’s the twist. Employers with 100 or fewer employees, who aren’t required to file federal EEO-1 reports, get a lighter recordkeeping burden under the Uniform Guidelines. They aren’t required to formally calculate adverse impact or keep candidate data broken out job by job the way larger employers must.

That sounds like a pass. The guidelines say otherwise, in as many words: small employers remain fully covered by federal equal employment opportunity law. The paperwork requirement is lighter. The underlying obligation not to discriminate is not.

Put the two pieces together and you get the real picture for a business your size: the four-fifths ratio was never going to be a reliable safety net, because your numbers are usually too thin for it to mean much either way, and you’re not even required to calculate it in the first place. Nobody’s checking your math but you. That was never a guarantee you’re clear.

Does that mean small businesses shouldn’t use assessments?

No. It means you should be honest about what an assessment does and doesn’t do for you.

What a structured assessment actually fixes

An unstructured gut-feel interview has its own well-documented consistency problem: different interviewers weighing different things, for different candidates, with nothing written down about why. That risk doesn’t disappear just because nobody measures it. It’s just invisible, which is worse, not better.

A structured assessment applied the same way to every candidate for a role removes a real source of that inconsistency. That’s a genuine improvement over deciding on vibes. It is not the same as eliminating adverse impact, and you should be skeptical of anyone, including us, who implies otherwise. No assessment, no vendor, no algorithm can honestly promise a discrimination-proof process. What a well-built one can do is apply your own defined criteria consistently and show its reasoning, so a problem has somewhere to get caught instead of nowhere.

What you’re actually buying, assessment by assessment

It’s also worth knowing what you’re actually buying. A personality assessment built on the IPIP Big Five model is a validated instrument with published reliability data. A situational judgment test is not a validated psychometric tool. It measures how closely a candidate’s approach lines up with how your team says it wants situations handled, which is exactly why one SJT’s answer key can look completely different from another’s. An environment fit assessment works the same way: it surfaces preference alignment, not a scientifically proven prediction of who’ll stay. None of the three hands you a verdict. All three hand you evidence to weigh, which is the same standard we walk through in more depth in how to build hiring assessments that actually hold up.

What a defensible assessment process actually looks like

The goal isn’t a perfect ratio. It’s a process you could explain, criterion by criterion, if a candidate or a regulator ever asked you to.

Set the criteria before you see a candidate

Start by defining what the role actually needs before you look at a single candidate. Must-haves, deal-breakers, and what “good” looks like for this specific job, not a generic template someone else wrote. Then run every candidate for that role through the same assessment, scored the same way. Consistency is the part you actually control.

This is the shape we built Truffle’s assessments around. Truffle is a candidate screening platform that combines resume screening, one-way video interviews, and talent assessments, and the Personality, Situational Judgment, and Environment Fit assessments run against the criteria you set for the role, not a hidden formula built into the product. You see why a candidate scored the way they did, not just the number. A gap between what you asked for and what a candidate showed becomes something to ask about in the next conversation, not an automatic disqualifier. AI surfaces the alignment. You still decide who moves forward.

Watch your own shortlists over time

Then keep an eye on your own shortlists over time. If your top scorers for a role keep skewing the same way demographically across several hiring rounds, that’s worth a second look, whether or not any single round would trip a formal ratio. You don’t need a compliance department to do this. You need to actually look. Pricing starts at $49 a month with a 7-day free trial, and every assessment ships on every plan, so this isn’t a tool you have to grow into before you can use it responsibly.

Adverse impact isn’t a box you check once

The instinct to treat adverse impact as something you calculate, clear, and file away is understandable. It’s also the wrong mental model for a business your size, where the math is often too thin to trust and nobody’s required to run it for you anyway.

The posture that actually holds up is closer to how you’d want any decision in your business reviewed: criteria set in advance, applied the same way to everyone, with the reasoning visible enough that you could stand behind it out loud. That standard doesn’t change whether the tool in question is a resume screen, a one-way interview, or an assessment. It’s worth revisiting each hiring season, not just the first time you turn a new tool on. If you want the fuller framework we use for evaluating any AI-assisted hiring tool against this bar, we’ve written more on responsible AI in hiring.

Frequently asked questions about adverse impact and hiring assessments

Does adverse impact only apply to companies with an HR department?

No. Title VII covers any employer with 15 or more employees for at least 20 weeks in the current or prior calendar year, regardless of whether they have an HR team. Many states set an even lower threshold, sometimes as low as three or five employees, so a small business can be covered well before it has anyone dedicated to hiring.

What’s the four-fifths rule, in plain terms?

It’s a rule of thumb the EEOC uses: if one group’s pass rate on a selection step is less than 80% of the highest-scoring group’s rate, that’s generally treated as evidence of adverse impact worth investigating. It isn’t a legal verdict on its own, and the government’s own guidance notes it becomes unreliable with small candidate numbers, in both directions.

Are personality tests, situational judgment tests, and environment fit assessments equally risky?

They’re different tools with different evidence behind them. A personality assessment built on the Big Five model is a validated instrument. Situational judgment tests and environment fit assessments are not validated psychometric tools. All three should be applied consistently and treated as evidence for a human decision, not a pass or fail verdict.

Can a hiring assessment protect my business from a discrimination claim?

Not by itself, and be wary of anyone who tells you otherwise. What an assessment can do is replace inconsistent, undocumented gut-feel decisions with a criterion-based process you can explain and defend. That reduces one real source of risk. It doesn’t eliminate your legal exposure, and no honest vendor will claim it does.