Comparison
Savvanta vs Codility
Codility is a mature technical-screening platform with real proctoring capability. The difference that matters is not whether it can watch a candidate — it can — but that watching is a setting the employer chooses per assessment, and in Savvanta it is a condition of the assessment existing at all.
Is the candidate proctored?
Optionally, and under employer control. Video proctoring is a premium feature bought separately through a customer success manager, it is disabled by default, and it is set per test. It also cannot be switched on for a test that already has candidates invited — the test has to be duplicated first, so proctoring cannot be added part-way through a live campaign.
Webcam, microphone and a full-monitor screen share are collected in one gate before the attempt exists. Deny any of them and the assessment does not start and the clock does not run. There is no reduced-proctoring path and no employer toggle that turns it off.
What this meansAn optional control produces an inconsistent record. When proctoring is off for one candidate and on for another, the two results are not comparable, and nobody notices until the decision is challenged.
What is the integrity signal made of?
Behavioural event detection — timestamped tab switches and paste flags — plus similarity checking on submitted code, and whatever screen or camera capture was purchased and enabled.
Source: Codility Support — Proctoring: behavioral events detection
One red-flag class only — the candidate left the assessment window. Face presence is measured but reported as a duration rather than a flag, and it carries no penalty. Recording interruptions draw down a 90-second budget and are reported as capture health, separately from performance.
What this meansWe deliberately produce fewer signals and refuse to grade most of them. An integrity log that renders a verdict invites the reader to treat a bandwidth problem as dishonesty, which is the most common way these systems hurt the wrong person.
What does the candidate know beforehand?
The invitation states the task and time limit. Marking criteria are not published to candidates in advance.
Every enforced requirement is published before the attempt starts, each labelled with whether it is enforced in the browser, on the server, or both — along with the device requirement, the real time limit including any granted adjustment, and how to request one. The mark scheme is public on this website.
What this meansTelling candidates the rubric does not make the work easier. It makes the result mean something, and it removes the single most common ground for a complaint that lands on your desk months later.
What Codility does better
- Similarity and AI-pattern checking on code against a large body of prior submissions.
- Depth of technical library and calibration data across many years of engineering screens.
- Enterprise workflow — ATS integrations, bulk invitations and reporting built for large technical hiring programmes.
- Flexibility. Some employers genuinely want light-touch proctoring for a first screen, and Codility lets them have it. We do not.
Read the thing itself rather than a claim about it. The report an employer receives is published in full — real criteria, real weights, an invented candidate, and the conduct finding placed above the score. No form.
Read a sample reportOr browse the 64 mark schemesFrequently asked questions
Does Codility record the screen?
It can. Screen recording and camera-plus-audio recording are part of Codility's Video Proctoring, which is a premium feature purchased separately, disabled by default, and enabled per test. In Savvanta it is not a setting and not an upsell — the assessment does not start without it.
Why is your proctoring not optional?
Because comparability is the point. If two candidates for the same job sat under different observation, the scores are not on the same scale, and that is the first thing an employment lawyer or a works council asks about.
Isn't mandatory recording harsh on candidates?
It is intrusive and we do not pretend otherwise, which is why every requirement is published before the attempt begins, adjustments are a first-class setting, and there is an appeal with written reasons. See the candidates page.
Do you check code for plagiarism?
Not against a large historical corpus in the way Codility does. Our answer is the recording and the follow-up explanation rather than statistical similarity.