Assessment design
Extra Time Has No Statutory Multiplier. Judge the Work, Not the Rate
No US employment statute, regulation or agency document sets an amount of extra time as the disability accommodation on a timed hiring assessment. The obligation is individual: what the candidate in front of you needs to show the same capability, which usually means asking them rather than picking a percentage on their behalf. Comparability only breaks if the clock was part of what you were measuring, so settle that first. If speed is not the construct, drop the limit for everyone and the request disappears.
The takeMost hiring time limits exist to protect the reviewers, not to measure the candidate. They cap how long a grader spends and they keep submissions uniform enough to skim. That is a real operational need and a bad reason to put a clock in front of a person, because the moment somebody asks for more of it you have to defend the limit as a measurement and you cannot. Fix the grading load somewhere it does not change what the assessment means.
Where Olive fits
Open a role and see what the work shows
If you are building the timed version yourself, the expensive parts are the answer key and the evidence behind each finding. Olive runs a role-grounded assignment, 40 to 60 minutes on the candidate's own clock, and returns six findings anchored to timestamped excerpts from the session, in a report the candidate is granted in identical form.
Rank your shortlistHow much extra time does the law actually require?
As much as this candidate needs to show the capability, decided one person at a time. The live regulation, 29 CFR 1630.11, says a test given to someone whose disability impairs sensory, manual or speaking skills has to be administered so the results reflect what the test claims to measure rather than the impairment 1. It names no format, no timer and no percentage.
That section governs administration only. The general duty to accommodate a known limitation sits at 29 CFR 1630.9, it turns on what this person needs, and its only stated limit is undue hardship 1. The exception inside 1630.11 is the whole of the comparability question: the administration duty does not apply where the impaired skill is the factor the test purports to measure 1. The EEOC's 2022 technical assistance names extended time and an alternative version of the test, including one compatible with a screen reader, as accommodations that may be effective, and it attaches no quantity to either 2. That document was removed from eeoc.gov in January 2025 and is read here from an archived capture, so use it as reasoning rather than as authority.
Half again as much time is a convention that travels between HR posts. Quoting it to a candidate carries two costs: it misstates the law, and it answers a question they did not ask. Most requests are narrower and more concrete than a multiplier. Somebody asks for a screen reader that works with the interface, a way to pause between sections, the instructions in writing, or a longer window to work inside. Ask what is being requested before deciding what to grant. The rest of that conversation is what ADA accommodations mean for an AI-based assessment.
Decide whether speed is part of the job
Open the assessment brief and write one sentence naming what the time limit measures. Answer that before granting anything, because it decides everything downstream. If the sentence is about the job, the limit is part of the construct and you can defend it. If it is about keeping grading volume down, or about stopping candidates from using an assistant, the limit is administrative and it should come off for everyone.
The same test appears in the accessibility standard, which is worth reading even though it binds almost nobody in private hiring. The Timing Adjustable criterion in WCAG 2.1, Success Criterion 2.2.1, requires a time limit to be turnable off, adjustable over a wide range, or extendable, unless the limit is essential and extending it would invalidate the activity 3. That essential exception is exactly the claim your one sentence has to make. WCAG carries no force of law by itself. Two US rules adopt it: WCAG 2.1 Level AA for state and local government web content, under the Department of Justice's Title II rule published on 24 April 2024, and WCAG 2.0 Level A and AA for federal agencies, under the Revised Section 508 Standards of January 2017. A private employer conforms only by choosing to.
Speed is genuinely the construct in some roles. A trading desk, an emergency dispatcher, a support queue with a hard first-response target: there the clock is the work. Say so in the invitation, name the reason, and hold the line. What breaks defensibility is not a timed exercise; it is a timed exercise whose timing nobody in the company can explain. Whether the round should carry a clock at all is the wider question in what a time limit on an assessment measures.
Does an extended clock make the results incomparable?
Only if the clock was part of the measurement. A timed assessment converts an artifact into a rate, and a rate stops being comparable the moment one person's denominator differs. Judge the work instead: what the candidate produced, what they chose to check, what they refused to accept from an assistant. None of that changes when one session ran ninety minutes and the next ran sixty.
What an assessment is about does more work than what format it takes, and one comparison sizes that. Inside the job knowledge meta-analysis Sackett and colleagues rely on, all 164 studies together produced a mean observed validity of .22, while the 59 studies using tests built for the job in question produced .31, rising to .40 once corrected for unreliable performance ratings 4. Those are two subsets of a single meta-analysis and nobody ran an experiment, and job knowledge tests assume candidates who already have the knowledge. Nothing in it measured a timer, so carrying the point across to clocks is inference. What the numbers support is narrower and still useful: relevance is where the signal was found.
What does have to stay fixed is everything that is the measurement:
- The same task, at the same difficulty, for every candidate in the role.
- The same rubric, written before anyone submits, with the anchors settled.
- The same evidence, meaning the reviewer looks for the same kinds of moments in every session.
- The same reviewer discipline, so nobody grades an extended session more sceptically for knowing it was extended.
Extensions go wrong at that last bullet. A reviewer who knows a candidate had longer starts reading the work for traces of the accommodation, which is a bias introduced by the accommodation process rather than by the accommodation.
Write the invitation so the clock is defensible
Put four things in the message that goes out with the assessment, before anybody starts. What the exercise is measuring, in one sentence. How long it takes, and whether that duration is a limit or an estimate. That the timing can flex. And who to write to for anything beyond it, by name, with a line saying that asking changes the format and nothing about the evaluation.
The difference between a limit and an estimate is worth labouring, because candidates cannot tell them apart and will assume the harsher one. "Most people finish in about an hour, and the window is open for three days" is a different instrument from "you have sixty minutes", even when the median submission is identical. The first costs nothing and removes a whole category of request before it is made. Designing the default so the request becomes unnecessary is the substance of offering accommodations without forcing a disclosure.
Then hold the record separately. Note what was requested, what was provided, what the reviewer saw and the date, and keep it out of the file the hiring panel reads. The obligation is individual, so the record is individual. The version of this that fails under scrutiny is the one where somebody applied a percentage found in a blog post to a person they never spoke to. What your own process has to grant, and what it has to record, are questions to confirm with counsel before the invitation goes out.
Common questions
Is fifty percent extra time the standard?
There is no standard. No US employment statute, no regulation and no agency document names a multiplier, and the figure people repeat comes from convention rather than from law. What the ADA's implementing regulation requires is that a test be administered so the results reflect what it claims to measure rather than an impairment 1, which is an individual question. Ask the candidate what they need, grant what is effective, and record the decision. Quoting a percentage to someone as though it were the rule invites a correction you will not enjoy.
Do you have to grant extra time on an untimed take-home?
There is usually nothing to grant, which is the point of an untimed take-home. If the exercise has an open window and a realistic effort estimate, the most common timing request disappears before anybody has to make it. What can still arrive is a request about the window itself, when somebody needs to work in shorter sessions or around an appointment. Extending a deadline you set for scheduling convenience costs nothing and changes no part of the measurement.
Can you tell the reviewer that a candidate had extra time?
Prefer not to. If the clock is not part of what the exercise measures, the reviewer has no use for the information and every opportunity to be swayed by it. Keep the accommodation record with whoever administers the process, not in the packet the panel reads. The exception is a case where the reviewer genuinely needs to know the format differed, such as a session delivered in writing rather than aloud, and then tell them the format and not the reason.
What if speed genuinely is part of the job?
Then say so, in the invitation, with the reason attached. A dispatcher, a trading desk and a first-response support queue all measure work against a clock, and an assessment can too. The ADA's test-administration duty carries an express exception where the impaired skill is the factor the test purports to measure 1. What that exception will not cover is a timer inherited from a template, so write down what the limit measures before you rely on it, and be ready to say it out loud to a candidate.
Does an accommodation on the assessment carry into the job?
No. Accommodation for the application process is its own question, decided against what the process demands, and accommodation to perform the role is decided later against the essential functions of that role. Granting an alternative format for an exercise commits you to nothing about the job, and it is not an admission that the role needs adjusting. Treat the two conversations as two conversations, and do not let the first one leak into a hiring decision.
References
- 1. 29 CFR 1630.11 - Administration of tests (Regulations to Implement the Equal Employment Provisions of the Americans with Disabilities Act) govinfo.gov Supports the test-administration duty at 1630.11 and its exception where the impaired skill is the factor the test purports to measure, and the separate reasonable-accommodation duty at 1630.9; the URL is the full text of Part 1630.
- 2. The Americans with Disabilities Act and the Use of Software, Algorithms, and Artificial Intelligence to Assess Job Applicants and Employees web.archive.org Supports the claim that extended time and a screen-reader-compatible alternative are named as possible accommodations with no quantity attached.
- 3. Web Content Accessibility Guidelines (WCAG) 2.1, Success Criterion 2.2.1 Timing Adjustable w3.org Supports the Timing Adjustable requirement and its essential exception, used here as the test a time limit has to pass.
- 4. Revisiting Meta-Analytic Estimates of Validity in Personnel Selection: Addressing Systematic Overcorrection for Restriction of Range static1.squarespace.com Supports the comparison between all job knowledge studies and the job-specific subset, used for the claim that content relevance carries more of the signal than administration does.
4 sources, numbered by first appearance. How Olive sources claims
General guidance for hiring teams. What works at one company and one volume may not transfer to yours.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.