Comparison
Savvanta vs TestGorilla
TestGorilla is a large, well-built library of pre-employment tests with a self-serve funnel and a decade of items behind it. The clearest way to explain the difference is to quote what TestGorilla itself recommends, and notice that it is not what TestGorilla sells.
TestGorilla's own conclusion. In a post titled “Why it's impossible to prevent cheating”, TestGorilla writes that no assessment “has ever been 100% cheat-proof”, and that its detection “doesn't and cannot tell you who cheated; it can point to behaviours which you can contextualise”. Its recommendation is to stop building walls and instead use live, real-time formats — simulations and spoken responses — where outsourcing the answer stops being practical. We agree completely. That recommendation describes what Savvanta is.
What is actually captured while the candidate works?
Webcam images at intervals rather than continuous video, with full-screen enforcement, tab-switch detection, paste tracking, IP monitoring and developer-tools detection. The screen itself is not recorded. Findings are summarised into three behaviour tiers — no issues, some issues, major issues.
Source: TestGorilla Help Centre — Understanding anti-cheating measures and behavior tiers
Screen, webcam, microphone and — on Windows and ChromeOS — system audio are recorded continuously for the whole attempt, chunked and stored, not sampled. Stop any required stream mid-attempt and a 90-second budget starts draining across the rest of the test; at zero the attempt is red-flagged and submitted for you.
What this meansA photograph every half-minute establishes that a face was present. It cannot show what the hands were doing in between. If the artefact you want is “show me the work happening”, only one of these two produces it.
Is proctoring guaranteed, or a setting?
Anti-cheating settings are configured per assessment by the employer, and the heavier capabilities — including ID verification — sit in the higher paid tiers rather than the self-serve plan.
Source: TestGorilla Help Centre — Understanding anti-cheating measures and behavior tiers
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 meansWith a toggle, two candidates for the same job can be assessed under different conditions and compared as though they were not. A gate removes that question, which matters more for defensibility than for detection.
What does the employer receive at the end?
Scores per test with percentile benchmarking against their test-taker population, plus the behaviour tier. The item bank and its scoring are not published to buyers before purchase.
A report against a six-criterion scheme published on this website in advance, with the timestamped evidence quoted beside each score, conduct findings placed above the score rather than averaged into it, and a named human required to confirm before it counts. Read one in full — no form.
What this meansYou can read our marking instrument and a finished report before you speak to anyone. Across sixteen assessment vendors we surveyed, none publish either.
What happens when a candidate says the result is wrong?
Handled between the employer and candidate. There is no published candidate-facing appeal process with a written-reasons requirement.
Seven appeal grounds — including proctoring flags, recording failure, platform faults, an adjustment not applied, the time limit, and the grade itself. A reviewer must write reasons, and the candidate sees them. Extra-time adjustments are a stored setting on the candidate, not an email to support.
What this meansIf you hire in the EU, or anywhere a declined candidate can ask a regulator how the decision was made, this is the part that gets asked about.
What TestGorilla does better
- Breadth. Hundreds of validated tests across cognitive ability, personality, software and language, available immediately.
- Norms. A large test-taker population means percentile benchmarks we cannot match. If you need to say "this candidate is in the 80th percentile of people who took this exact test", buy theirs.
- Self-serve and price. You can start today with a card. Our work needs a conversation first.
- Top-of-funnel throughput. Cheap per candidate, so it scales to a volume where a recorded human-confirmed assessment does not make sense.
The sane combination for a high-volume funnel is both: a cheap library test to cut the field, and a recorded role-specific assessment on the survivors.
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 TestGorilla record the candidate's screen?
No. TestGorilla's anti-cheating measures capture webcam images at intervals along with full-screen enforcement, tab-switch and paste detection; the screen itself is not recorded. Savvanta records screen, webcam and microphone continuously for the whole attempt.
Is Savvanta cheaper than TestGorilla?
No, and not by a small margin. TestGorilla is a per-candidate library product and Savvanta is a configured, recorded, human-confirmed assessment. They are priced like the different things they are.
Can I use both?
Yes, and for a funnel above a few hundred applicants it is the arrangement we would recommend even though it means paying someone else first.
Do you claim to stop cheating?
No. TestGorilla is right that no assessment is cheat-proof and we say the same thing on our own fraud page. What we claim is that the work is recorded, the task is hard to outsource, and the candidate has to explain it afterwards.