What kinds of behaviour do proctored online coding assessments typically flag for review?
answer
- The rules screen is part of the test
- Platforms count, humans judge
- Focus loss, pastes, similarity
- Permitted tools differ per employer
- A flag is a prompt, not a verdict
basics
~20 sCommon flags include leaving the assessment window, large paste events, abrupt jumps in typing activity, and similarity between your submission and other code. Which signals are collected varies by platform and by what the employer switches on.
solid answer
~50 sMost proctored assessment platforms attach a small set of counters to the score report rather than making a judgement themselves: how often the assessment window lost focus, how many paste events occurred and how large they were, unusual gaps or bursts in typing, and a similarity comparison against other submissions and widely published solutions. Camera or screen capture appears only where the platform supports it and the employer enables it, with a consent notice at the start; requirements around that vary by jurisdiction, so read the notice rather than assume. The practical move is to read the rules screen before starting, because whether an external editor, documentation or a search engine is allowed differs by employer, and then to stay inside those rules openly. A flag is a prompt for a human to look, not an automatic verdict.
go deeper
Know that these assessments record simple signals and that the rules about editors, documentation and search differ by employer. Read the instructions screen before the countdown starts rather than guessing.
Be able to describe the mechanics: focus-loss counts, paste sizes, editing playback and similarity comparison, all attached to the report for a person to interpret, with camera capture only where enabled and consented.
Show judgment about context — why an innocent paste from a permitted editor still gets logged, why a pattern rather than a single counter draws scrutiny, and how to explain your workflow plainly without sounding defensive.
Own the tradeoff in requiring proctoring at all: it deters the obvious cases and produces noisy counters that a reviewer must interpret, and heavy-handed settings cost you candidates. Decide deliberately what your screen needs rather than accepting platform defaults.
## What proctoring means on an assessment platform Proctoring here rarely means a live invigilator. It usually means instrumentation: the assessment page records a handful of signals while you work and attaches them to the score report the employer receives. The platform generally does not decide anything. It counts things, and a human decides what the counts mean. The signals that appear most often: - **Window or tab focus loss** — a count, and sometimes durations, of times the assessment page stopped being the active window. - **Paste events** — how many, and how large. Some platforms record the pasted length; some record the content. - **Typing rhythm** — long silences followed by a large block of finished code look different from incremental authoring, and playback of the editing timeline is a feature on several platforms. - **Similarity** — an automated comparison of your submission against other candidates' submissions for the same problem and against widely published solutions. - **Camera or screen capture** — present only where the platform offers it and the employer turns it on, always behind a consent screen at the start. What is permitted and what notice is required varies by jurisdiction and by employer policy, so treat the notice you are shown as the authority for that assessment rather than assuming a universal rule. ## Read the rules screen; it is part of the test The most common self-inflicted problem is not cheating but guessing. Norms genuinely differ: one employer allows an external editor and open documentation, another allows documentation but not an editor, another asks you to stay in the browser tab entirely. That is stated on the instructions or consent screen, which many candidates click through in a hurry because the countdown is visible. Worth checking before the timer starts, on the invented assessment used through this leaf — a mid-size product consultancy running three problems in an 82-minute window for junior frontend candidates — the rules screen answered all of the following: whether an external editor was permitted, whether the window was global or per problem, whether the timer began at the invite link or at the first problem, and whether documentation lookups counted as leaving the assessment. ## Innocent behaviour that still reads as a flag Counters have no context. If the rules allow you to draft in your own editor, the paste event that brings your own code back into the browser is logged exactly like any other paste. That is not a reason to avoid a permitted workflow, but it is a reason to know the flag exists and to be able to explain it plainly if asked: the code is yours, the workflow was permitted, and the editing history supports it. Two focus losses to check documentation the rules allow are ordinarily unremarkable. What draws scrutiny is a pattern — repeated focus loss timed against large pastes, a submission arriving with no incremental editing history, and a high similarity score together. ## What actually happens when something is flagged Practice varies by employer and is worth stating carefully: many treat a confirmed integrity failure as ending the process for that role, and some record it against future applications on the same platform or with the same employer. Others ask the candidate to explain, or re-test with a live interviewer. What is fairly consistent is that a raw counter alone does not usually end anything; a similarity match against another candidate's near-identical submission is the signal that reliably does. ## The candidate's version of the rule Do permitted things openly and impermissible things not at all. Do not paste in a solution you found; do not have a second person in the room or on the call; do not run the assessment for someone else. Beyond that, the defensible position is simple and does not require guessing at the platform's internals: you followed the rules on the screen you were shown, and the editing history of your submission looks like someone solving a problem, because that is what happened.
- If the rules allow an external editor, how do you keep a large paste from reading as cheating?Accept that the paste will be logged and make it explainable. Keep the workflow to your own code, build the solution incrementally rather than arriving with a finished block, and if the platform offers a notes field or the recruiter asks, say plainly that you drafted in a permitted external editor. Counters are read by a person in context, and a consistent editing history supports the account.
- A candidate loses window focus twice to check documentation the rules permit — is that a problem?Ordinarily not on its own. Two focus losses inside a long window, where documentation was explicitly allowed, is close to background noise. What draws attention is a pattern that lines up with something else — repeated focus loss immediately before large pastes, or a high similarity score on the same submission. Practice differs by employer, so the rules screen remains the authority.
- What should you check on the instructions screen before the countdown starts?Which tools and references are permitted, whether the timer starts at the invite link or the first problem, whether the window is global or per problem, whether you can move between problems freely, and what happens if the connection drops. Those five answers change how you plan the window, and all of them are usually on the screen candidates skim fastest.
saying these in an interview costs you the question
- Starting the assessment without reading the rules or consent screen
- Assuming any tool is allowed unless explicitly banned
- Pasting in a solution found elsewhere during the assessment
- Having another person present or screen-sharing during the window
- Believing every logged counter is an automatic rejection
- Assuming proctoring settings are identical across every employer