AI assessment templates
A template defines what a session covers and how long it runs. One template is reused for every candidate you invite to that role.
Settings
| Setting | Default | Notes |
|---|---|---|
| Technology | — | What the assessment is about, e.g. Go, React, PostgreSQL |
| Focus areas | — | Named skill areas, each with a weight |
| Maximum questions | 15 | The session cannot run longer than this |
| Time limit | 30 minutes | Whichever limit is reached first ends the session |
Templates start as drafts and are activated when you are ready to invite people.
Focus areas and weights
Focus areas are the skills within the technology — for React that might be hooks, state management, rendering behaviour, testing. Each carries a weight.
The weight decides how many questions that area gets, and question selection continually picks the area furthest behind its target. So the weights are not decoration: they are the specification of what the assessment is actually measuring.
Two things follow:
Weight what the job needs, not what is easy to ask. It is tempting to weight the areas with the crispest questions. The result is a report about those areas, presented as a report about the technology.
Do not spread fifteen questions over eight areas. Two questions per area cannot distinguish someone competent from someone lucky, and the skill breakdown in the report will invite exactly that reading. Three or four areas at fifteen questions is a sensible ceiling.
Session length
Maximum questions and time limit are both caps, and the session ends at whichever comes first.
Consider what happens when the clock runs out mid-session: coverage will be uneven, and the areas selected last are the ones that get missed. If you want fifteen questions answered, allow enough time that a thoughtful candidate reaches fifteen — a rushed session and a short session produce differently biased reports.
Inviting candidates
Invitations work like assessment invitations: one per candidate email, with a token link, an expiry, and an attempt allowance defaulting to one.
Candidates need no account. They open the link, declare their level, and start.
Changing a template
Editing changes future sessions only; completed sessions keep the template they ran against.
Changing focus-area weights makes candidates non-comparable, and nothing on the report will tell you that. Two people with the same score assessed under different weightings were measured against different specifications. If you change weights mid-funnel, note when you did it.