Assessments3 min read
Test integrity without the paranoia
Remote tests need some protection. The fair version records openly, raises flags for a person to review, and never turns a flag into a verdict.
Assessments3 min read
Remote tests need some protection. The fair version records openly, raises flags for a person to review, and never turns a flag into a verdict.
Put this into practice on your next role.
Free for your first role. No card.
Design tests where looking something up does not give the answer away.
Once a test is taken at home, someone will ask how you know the candidate did it alone. It is a fair question. The answers on offer range from sensible to invasive, and the invasive ones tend to catch nervous people rather than dishonest ones.
Good integrity practice starts from a simple position: most candidates are honest, the ones who are not are rarely stopped by surveillance alone, and every measure you add costs the honest majority something.
The strongest protection is a test that is hard to game. Questions that ask someone to reason about a specific scenario, explain a choice or write working code are much harder to look up than questions with a single memorised answer. A short live follow-up, “walk me through how you got there”, makes outside help obvious very quickly.
If you record a sitting, say so before it begins, in plain words: what is captured (screen, camera, microphone), who can see it, and how long it is kept. A candidate who knows what is happening performs normally. A candidate who discovers it afterwards stops trusting everything else you told them.
Automated checks are useful for one thing: pointing a person at a moment worth reviewing. The candidate left the tab. A long answer was pasted in at once. Nobody was in front of the camera for a minute.
Every one of those has an innocent explanation. A notification pulled focus. The candidate drafted an answer in a text editor first. They stepped away to deal with a child or a power cut. Treat a flag as the start of a question, not the end of one.
A flag tells you where to look. It never tells you what you will find.
Look at the moment
Watch the recording around the flag. Most resolve in seconds.
Weigh it against the whole
One tab switch in a forty-minute test means little. A pattern across every hard question means more.
Ask, if it matters
If a flag would change your decision, ask the candidate about it. People can usually explain, and how they explain is informative too.
?Quick check
A recording shows a candidate left the test tab three times in forty minutes. What do you do?
Many candidates sit tests on a shared laptop, on a patchy connection, in a busy home. Build in slack: allow a reconnect, let them resume where they left off, and give a window of days rather than a fixed slot. A test that only works in perfect conditions measures the conditions.
Up next
Skills tests that candidates don’t resent
A good test is short, clearly about the job, and honest about what it measures. A bad one is unpaid work with a timer on it. Candidates can tell the difference immediately.
Read next · 3 minWas this useful?
One tap. It decides what we write next.
Get the next one by email
One short email when something new is worth reading.
A good test is short, clearly about the job, and honest about what it measures. A bad one is unpaid work with a timer on it. Candidates can tell the difference immediately.