The Pre-Hire Assessment Reference, One Row per Test Type You Get Sent

An assessment invite usually names a platform and a deadline and very little else. This reference is for candidates who have just been sent one. Find the row for your test, see what it is scored against, whether practice will move the result, which monitoring is switched on, and how to ask for adjustments before the clock starts. Where the public evidence is thin, I say so and tell you what to ask.

What does a pre-employment assessment test? One row per type

A pre-employment assessment measures one of a small set of things: judgement in work scenarios, reasoning ability, personality traits, integrity, a job skill such as typing or coding, or, in game-based tests, traits the employer often does not name. The US Office of Personnel Management (OPM) Assessment Decision Guide covers cognitive ability, integrity, job knowledge, personality, situational judgement and work-sample tests. Add typing and game-based tests and you have covered nearly every invite you will get.

"Not published per type" means exactly that. I found no public statement, test by test, of whether each type is scored against a fixed key, a norm group or a fit profile. Where the table says "inferred", treat it as a working guess and confirm it with the recruiter.

Situational judgement test (SJT)

A situational judgement test describes work scenarios and asks you to pick, rate or rank possible responses.

• Recognise it by: a scenario, then options you mark best and worst, or rank.
• Scored against: a key. OPM documents that "Scores are based on subject matter experts' judgments of the best and worst alternatives." Prep vendors say the raw score is then compared with a previous group of candidates.
• Prep payoff: real. Commercial coaching raised SJT scores by d = 0.59 in a high-stakes medical school admission setting.
• How it misleads: it reads like common sense, so candidates skip prep. The key reflects someone's judgement, and the employer's stated values probably shaped it.

Cognitive ability

A cognitive ability test measures reasoning: numerical, verbal or abstract problems with one right answer.

• Recognise it by: problem sets with a single correct option, usually against a countdown.
• Scored against: not published per type. Ask whether results are reported as a raw score or a percentile.
• Prep payoff: real but capped. Scores rose up to the third administration and then plateaued.
• How it misleads: candidates keep drilling long after the gains have stopped.

Self-report personality

A self-report personality test asks you to describe yourself, either by rating single statements or by picking which of several statements is most like you (forced choice).

• Recognise it by: agreement scales, or sets of statements where you choose one.
• Scored against: a trait profile, which I infer from the format. Nobody has published an official scoring model for "fit-to-incumbent" profiles, the kind that compare you with current staff.
• Prep payoff: not studied. The published evidence measures faking (deliberately answering to look good), not practice.
• How it misleads: people assume it is easy to game. Inflation varies by trait, and forced-choice formats resist it.

Integrity

An integrity test measures attitudes to honesty and rules. Overt tests ask about honesty directly. Personality-based tests measure related traits indirectly.

• Recognise it by: direct questions about theft, rule-breaking or honesty (overt), or questions that look like a personality inventory (personality-based).
• Scored against: not published per type.
• Prep payoff: coaching moved overt tests by d = 1.32 and personality-based tests by 0.36.
• How it misleads: the plain intent of overt questions makes the "right" answer feel obvious to every applicant.

Typing

• Scored against: a fixed key. OPM measures speed in net words per minute.
• How it misleads: because speed is the skill being measured, an employer can lawfully keep the time limit (see the accommodations section).

Job knowledge, work samples and coding tests

A work-sample test asks you to do a realistic slice of the job, such as a coding exercise on a platform like HackerRank.

• Scored against: not published per type.
• Prep payoff: I found no practice-effect study for these.
• How it misleads: this is where the coding-platform proctoring controls apply, and several of them run without a prompt.

Game-based

A game-based assessment replaces questions with puzzles or mini-games.

• Scored against: no official scoring model has been published.
• Prep payoff: unknown.
• How it misleads: it feels like play, but it is scoring traits you cannot see. Ask in writing what characteristics it measures. If you live in New York City and the game counts as an automated decision tool, the notice you are owed must name them.

How is each test scored: fixed key, norm group, or fit profile?

There are three scoring models. A fixed key marks each answer right or wrong against a set standard. A norm group compares your raw score with a previous group of candidates and usually reports a percentile. A fit profile compares your trait pattern with a target. Only SJT and typing scoring are documented in the public sources I could check. For every other type, you infer the model and confirm it.

A percentile is the share of the comparison group you scored above. A 70th percentile means you outscored 70% of that group, not that you got 70% of answers right.

The three ways an SJT gets its key

The SJT is the best-documented case, and it shows that "the right answer" is a design choice:

• Expert scoring: subject matter experts judge the best and worst options. This is OPM's documented method.
• Target scoring: the test author decides the correct answer.
• Consensual scoring: each option scores by the percentage of people who chose it.

After keying, prep vendors say your raw score is compared with a norm group. So an SJT is often both: an expert key turns your answers into a number, and the norm group turns that number into a rank.

How much one SJT score tells you

A meta-analysis put SJT retest reliability at r = .698. Retest reliability is the correlation between the same people's scores on two sittings, where 1.0 would mean an identical order both times. At about .7, one sitting is a reasonable estimate but not a fixed fact about you. That matters if you are offered a retake.

Inferring the model from the invite

These are cues I would read, not documented rules:

• Speed or accuracy stated as the measure (words per minute, items correct): a fixed key.
• "Percentile", "compared with other applicants", "benchmark": a norm group.
• "Fit", "profile", "match to role", "culture": a fit profile. Expect a trait-based instrument, and expect no published scoring model.
• No clue at all: ask the recruiter. The question costs nothing, and the answer changes how you prep.

Will practice move my result?

It depends sharply on the type. Cognitive test gains plateau by the third attempt, SJT coaching moved scores by d = 0.59, overt integrity tests move the most, and nobody has studied practice on personality tests, only faking. In these studies, d expresses a score shift in standard deviations, so d = 0.59 means scores moved by a little more than half a standard deviation.

Two patterns explain most of this table.

• The more obvious the "right" answer, the more the score moves. Overt integrity questions show their intent, and coaching moved them nearly four times as much as personality-based ones (1.32 against 0.36).
• Applicants inflate the traits they think employers want. Emotional stability inflated by 0.94 in real applicants, extraversion by 0.29, agreeableness by 0.19. Researchers attribute the gap to applicants seeing emotional stability as valued.

These numbers describe how far each instrument can move. They are not advice to fake. Warnings lowered personality scores by d = .23, and forced-choice formats were, on average, more resistant to faking. Many employers use both.

Prep budgets by type

• Cognitive ability: about two full practice runs. Because gains plateau by the third administration, more runs add little.
• SJT: study the employer's stated values and competencies, because those probably shaped the key. If Refolk tailored your resume to this posting, it has already pulled out the posting's stated requirements. Reread that list as your SJT prep sheet.
• Personality and integrity: learn the format. Know whether items are single statements or forced choice, then answer as yourself.
• Game-based and work samples: no evidence on practice either way. Spend your time asking what is measured.

Before you set any budget, decide whether the application deserves an evening at all. Refolk's fit score gives a quick read on how well you actually match the posting. A weak match is a reason to move fast and keep your prep time for stronger roles.

Is my online job assessment proctored?

Assume some monitoring is on even if you see no sign of it. Controls that need your permission, such as a camera or full-screen mode, show up on the launch screen. Others, such as HackerRank's copy-paste tracking, which runs by default, show no prompt at all. On HackerRank the employer switches each control on, so two invites on the same platform can be monitored very differently.

What the launch screen tells you

Proctored tests usually show one or more of these:

• a consent notice
• a webcam permission request
• a full-screen permission request
• a warning that tab switches are recorded
• a prompt to install a desktop app

On HackerRank specifically:

• Image Proctoring (webcam snapshots) is off by default.
• Proctor Mode must be switched on at both the company level and the test level.
• Copy-paste tracking is on by default for customer tests.
• Desktop App Mode switches on AI plagiarism detection and image analysis by default.

Sources disagree on tab-switch tracking. A HackerRank-hosted article says tab proctoring is disabled by default and has to be switched on. A later third-party article says HackerRank's Test Integrity page now lists it as a default and that "Plain Tab Proctoring leaves no visible signature." I cannot settle this from public documents. Treat tab tracking as possibly on and possibly invisible.

I found no documentation of room scans or ID matching on these hiring platforms. If a test asks for either, that is unusual enough to ask the recruiter about.

The launch screen shows you the controls that need permission, not the controls that are running.

When face detection misfires

A face-detection flag means the software lost your face, not that you did anything wrong. In a peer-reviewed study of Respondus Monitor, an education proctoring tool, detection rates tracked skin tone:

Two limits apply. The study covered students, not job candidates, and the darkest-skin group was small (n = 41). The way it fails still carries over: a 14-point detection gap means more "missing from frame" flags that have nothing to do with behaviour. Proctoring systems also flag people who look away from the screen or move in unusual ways, such as neurodivergent candidates who stim. I found no fallback that the assessment providers themselves recommend. So the practical defence is to ask whether a person reviews flags before any decision is made.

Hi [Name],

Thanks for the assessment link for the [Role] position. Before I start, could you confirm a few details so I can prepare properly?

1. What type of assessment is this (for example situational judgement, cognitive, personality, coding, or game-based)?
2. Are results scored against a fixed answer key, compared with other candidates, or matched to a role profile?
3. Which proctoring features are switched on (webcam, full-screen, tab-switch tracking, copy-paste tracking, desktop app)?
4. If the system flags anything, does a person review the flag before a decision is made?
5. Is the deadline of [date] firm?

Thanks,
[Your name]

How to request extra time or an alternative process

Send the request to the employer's recruiter or HR, not to the test platform, before you open any timed section, and get the agreed change confirmed in writing. In most cases the candidate has to start this, because the employer is only obliged to accommodate limitations it knows about.

United States (ADA)

• Who starts it: usually you. Employers must accommodate limitations "known to the employer".
• When: sources disagree. The EEOC says the ADA "does not require that an applicant inform an employer about the need for a reasonable accommodation at any particular time." The regulation's interpretive guidance says employers must provide alternative tests "upon advance request." In practice, ask before you start.
• Speed tests: an employer can keep the time limit if speed is one of the skills being measured. Typing is the obvious case.
• Documentation: the employer may ask for medical information when your disability or need for accommodation is not obvious.

United Kingdom

• Standard offer: the usual suggestion from employers is an extra 25%. That figure comes from education, where 25% is the most common extra-time arrangement for GCSE, AS and A level exams.
• When 25% falls short: Cambridge's careers service warns it "might not be enough extra time for a psychometric test". One assessment provider says candidates with more significant processing-speed difficulties may need 50% or more.
• Who asks: you, of the recruiting organisation. SHL says adjustments "need to be requested by the candidate themselves from the recruiting organization."

New York City: Local Law 144

Local Law 144 regulates automated employment decision tools (AEDTs), which are software tools that score or screen candidates.

• Notice: NYC residents must get notice at least 10 business days before the tool assesses them. The notice must name the characteristics assessed and explain how to request an alternative process.
• No guaranteed alternative: compliance guides quote the city's rules: "Nothing in this subchapter requires an employer or employment agency to provide an alternative selection process."
• Data: if the data type, source and retention policy are not on the employer's website, they must be provided on written request.
• Penalties: up to $500 for a first violation, and $500 to $1,500 for each one after that.

I took these points from secondary compliance guides, not the city's own rule text. Read the city's published rules before relying on any one point.

Elsewhere

I have no established guidance for the EU or other places. Ask the employer for its adjustment process in writing and check local disability law.

Subject: Assessment adjustment request: [Role], [your name]

Hi [Name],

I have received the [assessment name/platform] link for the [Role] position, due [date]. Before I start, I would like to request an adjustment.

I have [condition, or "a disability"] that affects [processing speed / reading on screen / sustained eye contact with a camera / other]. To take the assessment on an equal footing, I am requesting:
- [e.g. 50% extra time on timed sections]
- [e.g. no penalty for looking away from the camera, with human review of any proctoring flag]

I am happy to discuss what works for the format, and to provide documentation if you need it. I will hold off starting until the adjustment is confirmed in writing. Could you also confirm whether the deadline moves to allow time for this?

Thanks,
[Your name]

From invite to submit: the procedure

Work through these seven steps between the moment the link arrives and the moment you submit. No source gives timings for steps 1, 3 and 6. The five minutes I suggest for step 1 is my estimate, and arranging an adjustment can take the employer some time.

How this goes wrong

Most assessment mistakes come from treating an absence as evidence: no visible key taken to mean no prep, no camera taken to mean no monitoring, no legal duty taken to mean no recourse. Each failure below has a check you can run in a few minutes.

1. Calling an SJT "unprepareable." It looks like a common-sense test, so candidates go in cold. Coaching moved scores by d = 0.59, and the key encodes someone's judgement of best and worst. Check: does the invite or posting name the employer's values or competencies? Those probably shaped the key, so prepare against them.

2. Treating personality tests as easy to game. Inflation ranges from 0.19 to 0.94 depending on the trait, warnings reduce it, and forced-choice formats resist it. A candidate who inflates every trait on a forced-choice test is fighting the format. Check: are the items single statements on a scale, or sets where you pick the one most like you?

3. Reading "no camera prompt" as "unmonitored." Copy-paste tracking runs by default on HackerRank, and tab tracking may leave no visible trace. A clean launch screen tells you only which controls need permission. Check: ask the recruiter which mode is on.

4. Treating a face-detection flag as evidence of cheating. Detection gaps follow skin tone (78% against 92%), and looking away or stimming gets flagged. A flag can come from the software failing, not from anything you did. Check: ask whether a person reviews flags before a decision, and add flag review to your adjustment request if it applies to you.

5. Assuming adjustments carry over from education. A school or university extra-time arrangement does not follow you into hiring, and 25% may not be enough for a psychometric test. Check: is the format limiting you, or the time? Ask for the adjustment that fixes that specific limit, which may be 50% or more extra time or a different format.

6. Assuming Local Law 144 gives you a human reviewer. The notice must explain how to request an alternative process, but the employer does not have to provide one. Check: other duties still apply separately. A disability accommodation request under the ADA is a different route with different obligations.

7. Asking the platform for adjustments. SHL says it is not involved in disclosure. A request sent to the test provider's support desk can sit unanswered while your deadline runs. Check: send it to the employer and copy the recruiter who sent the link.

8. Over-practising cognitive tests. Scores plateau by the third administration. An evening spent on a fifth practice run would do more good on another application. Check: count your full practice runs and stop at two.

Who builds these tests, and what that tells you

In Refolk's index of professional profiles, people with formal test-design titles are few, and they cluster in government and the public sector. The much larger group that lists psychometrics as a skill sits with founders and assessment firms. An industrial-organisational (I/O) psychologist is the US title for a psychologist who studies work and selection, and occupational psychologist is the UK equivalent.

The top US employers of titled I/O psychologists are DCS Corp, the SEC, the US Air Force and the US Navy. In the UK they are the College of Policing, the Ministry of Justice, Work Psychology Group and Strathclyde Fire & Rescue. Among US profiles listing Psychometrics as a skill, the most common titles are founder or CEO, and Pinsight is the most common employer. The UK has 1.11 times as many titled psychologists as the US.

Treat this as directional. The employer and title breakdowns come from a sample of 25 profiles per query. Here is how I would use it:

• Public-sector invites, especially in defence, policing and justice, are more likely to come from teams with titled psychologists and documented methods such as OPM's expert key for SJTs. Expect a defined key and ask for the competency framework.
• Invites from startups and newer assessment firms, including game-based tools, come from a much larger and more varied pool of builders. Expect the scoring model to be documented less often, and put your effort into asking what the test measures.

Before you click start, and keeping this reference current

Run the checklist below once per invite. Then re-check the parts of this reference that change: platform defaults, which can shift without notice, and the city rules behind Local Law 144.

What to re-check, and how

• Platform defaults. The tab-tracking dispute shows that defaults change and that secondary sources lag behind them. Before relying on any default in this reference, read the platform's current support documentation for its proctoring and integrity settings.
• Local Law 144 detail. Everything here comes from compliance guides. Read the city's published rules for the current notice, data-request and penalty terms.
• Disability guidance. The EEOC's guidance on job applicants and the ADA, and the regulation at 29 CFR 1630, are the primary US texts. For the UK, the careers service guides from universities and the assessment providers' candidate pages show current practice on extra time.
• Your own record. After each assessment, note the type, the platform, the visible controls and what the recruiter told you. A few entries will show you which employers in your target set explain their tests and which do not.