Before any interviewer sees your name, a coding assessment test usually sees it first. Which platform runs that test changes almost everything: how long you get, whether a human is watching, what counts as cheating, and whether you can retake it.
Almost every guide to these platforms is written for the wrong reader. Search the platform names and you get vendor comparison pages aimed at recruiters choosing software, or blogs funded by tools promising to beat the proctoring.
This page is for the person sitting the coding assessment test. Everything below is sourced to the platforms’ own documentation, and where we could not verify something we say so.
Two Corrections Worth More Than Any Question List
1. CodeSignal’s 600–850 scale no longer exists
You will read everywhere that you need “700+” on a CodeSignal assessment. That advice refers to a scale CodeSignal has retired.
The current Assessment Score runs 200 to 600. CodeSignal publishes its own conversion table, and the numbers are not intuitive:
- Old 600 → new 203
- Old 700 → new 389
- Old 800 → new 532
- Old 850 → new 600
So the famous “700” is roughly 389 today. If a guide tells you to aim for a score out of 850, it is describing a system that no longer runs.
One thing nobody can honestly tell you is the score a specific company requires. Employers set their own cut scores privately. Any page quoting “Company X needs 750” is inventing it.
2. The Karat redo does not replace your score
Karat lets you request a redo within 24 hours if you feel you underperformed. The widespread belief is that the second attempt replaces the first.
It does not. Per Karat’s own material, Karat publishes both recommendations to the client and recommends they review the higher of the two. That is a recommendation, not a guarantee — the employer sees both and decides.
Worth requesting anyway. Karat reports roughly 1,000 hires that came through redos. Just go in knowing the first attempt does not vanish.
Karat: A Coding Assessment Test Run by Strangers
Karat is unusual because your interviewer does not work for the company you applied to. Karat employs its own Interview Engineers who conduct the session on the employer’s behalf.
Format: Karat’s support documentation says 45–60 minutes; its marketing page describes a flat 60 minutes split into a short introduction, around 10 minutes of discussion questions and roughly 40 minutes of programming in Karat’s browser IDE. Those two figures disagree slightly — plan for the longer one.
Everything is recorded. Video, audio, your code and a full keystroke playback all go to the employer. Karat states interviews are recorded regardless of whether you turn your camera on. Employers can view but not download the recording.
The scoring criterion that explains everything
Karat’s results report scores coding on completeness, optimality, time and space complexity, debugging — and handholding.
Read that last one again. How much help you needed is itself a scored criterion. This is why candidates so often describe the Karat interviewer as unresponsive, or “like talking to a wall.” It is not personality. It is structural: giving you hints costs you points.
Outcomes come as labels rather than a simple pass or fail — Do Not Pursue, Requires Further Review, Invite to Next Round, or Fast Track.
Two things candidates get wrong about Karat
You are allowed to look things up. Karat explicitly permits looking up documentation during the coding portion, “just as you would on the job.” AI tools are permitted for documentation lookup but not for generating solutions.
Memorising leaked questions is low-yield. Karat controls its content so candidates do not see the same material in a subsequent interview.
One more development worth knowing: in December 2025 Karat launched NextGen Interviews, a human-led but AI-enabled format built around multi-file projects with an integrated AI assistant, designed around problems a model cannot simply solve for you. Rollout by client is not public, so you may still meet the classic format — but most guides describe only the old one.
CodeSignal: Standardised, Reusable, and Easy to Damage
The General Coding Assessment is 4 questions in 70 minutes, all visible from the start and answerable in any order. Outside IDEs are not permitted. It supports 45 languages, drawing on what CodeSignal says are over 628,000 question variations.
Scores are reusable. You can share an existing result with a new company instead of retaking — a real advantage over every other platform here.
But retaking carries genuine risk. If you retake and share the new result, the company receives only the new score, even if it is lower. There is no “best of” protection.
And often you cannot retake at all. CodeSignal runs cooldown periods — commonly one attempt per 180 days, sometimes two or three depending on the assessment. Inside a cooldown, you must reshare an old score.
On integrity, CodeSignal’s Suspicion Score watches paste events, description copying, similarity against other submissions, and language switching combined with pasting. Notably its documentation does not describe webcam or tab-switch monitoring for this feature. A flag is explicitly not an automatic rejection — it is a prompt for the employer to look closer.
HackerRank: Baseline Monitoring You Cannot Turn Off
A HackerRank coding assessment test is usually asynchronous within an invitation window, mixing multiple choice with coding.
Employers choose between Secure Mode (full-screen enforced, copy-paste blocked, multiple monitors prevented, tab switches flagged), Proctor Mode (adds AI screenshot analysis, plagiarism detection and webcam anomaly detection), and a fully locked-down Desktop App Mode.
Here is the part most guides get wrong. Two things run regardless of mode: copy-paste tracking and tab proctoring. And the copy-paste tracking does not merely note that a paste happened — the pasted content itself appears in your test report.
Webcam proctoring applies only to proctored tests, and your invitation email tells you if yours is one.
Codility: Scoring Well Is What Triggers the Plagiarism Check
A Codility coding assessment test typically runs from half an hour to two hours as one continuous session. The timer covers the whole test, not individual tasks, and you cannot pause it. Sit down only when you can finish.
Now the counterintuitive part. Codility runs its Similarity Check on candidates who reach the passing score — or, where no passing score is set, on everyone scoring at least 40%. Doing well is what puts your submission under review.
It compares against more than 12 million previous assessments plus scraped public solutions, and detects obfuscation such as renamed variables or reformatted code.
Codility is candid that this produces false positives: some tasks are too simple to distinguish coincidence from copying, and a candidate taking Codility tests for several companies can flag against their own earlier submission. Codility advises employers to ask rather than assume.
To its credit, Codility is the most transparent platform here about monitoring — enabled tracking is disclosed on the intro page before you start.
CoderPad: Every Keystroke Is Replayable
CoderPad Interview is the live collaborative pad; CoderPad Screen is its take-home product.
The quirk almost nobody knows: CoderPad records every keystroke of every participant, and interviewers can replay the whole session afterwards from their pads list.
Deleting something does not erase it. A large block of code appearing instantly is obvious in playback without any paste detector. Playback covers code edits only — not audio or video — and shell commands do not appear.
CoderPad also publishes a free Interview Sandbox mirroring the real interface. Practising in it is the cheapest preparation on this page.
HireVue: Mostly Not a Coding Test, and No Longer Scanning Your Face
HireVue does offer coding assessments, but its centre of gravity is video interviews, game-based psychometrics and job simulations.
The claim you will still read everywhere — that HireVue analyses your facial expressions — is out of date by roughly six years. HireVue discontinued facial analysis; reporting places the technical change in March 2020 and the public announcement in January 2021. Assessment now centres on language and speech content rather than visual analysis.
For completeness: in March 2025 the ACLU of Colorado filed a complaint involving Intuit and HireVue concerning accessibility for a deaf candidate. HireVue’s chief executive responded that the complaint is without merit and that Intuit did not use a HireVue AI-based assessment. It is an unresolved allegation and we report it as one.
How to Prepare for Each Coding Assessment Test
- Karat: practise narrating your reasoning without asking for help. Documentation lookup is allowed — get fluent at it.
- CodeSignal: practise four problems in 70 minutes as one block. Do not retake casually.
- HackerRank: type everything. Assume tab switches and pasted content are visible.
- Codility: block out uninterrupted time. The clock does not stop.
- CoderPad: rehearse in the free Interview Sandbox first.
- HireVue: prepare structured spoken answers; the content of what you say is what is assessed.
Frequently Asked Questions
What score do I need to pass a coding assessment test?
No coding assessment test platform publishes company thresholds, because employers set them privately. Treat any specific number you read elsewhere as invented.
Can I use my own IDE?
Not on CodeSignal’s General Coding Assessment, which prohibits outside IDEs. Other platforms vary and their candidate documentation is often silent on it.
Does turning my camera off stop the recording?
Not at Karat. All interviews are recorded regardless of camera state.
Can I retake a coding assessment test?
CodeSignal allows retakes subject to cooldown periods, but the new score replaces the shared one even if lower. Karat offers a redo on request within 24 hours, with both results going to the employer.
Will a plagiarism flag automatically reject me?
Not by design. Both CodeSignal and Codility describe flags as prompts for human review, and Codility openly acknowledges false positives.
Sources
Every claim above comes from the platforms’ own candidate and support documentation rather than third-party summaries — including CodeSignal’s published conversion table for the retired scoring scale, which is where the 700-to-389 figure comes from.
Where a platform’s documentation contradicts itself, or where only candidate anecdotes exist, we have flagged it rather than smoothing it over. Policies change — check your invitation email and the platform’s current docs before your test.
Whichever coding assessment test you are facing, the same principle applies: find out which platform runs it before you decide how to prepare. Ask your recruiter — it is a reasonable question and the answer changes your tactics.
Next, see which companies use which format in our hub of coding interview questions by company, or find the guide for your job in interview questions by role.
