BARS rubrics in minutes, not workshops.
Behaviourally-anchored rating scales are the gold standard panels rarely get — because writing good anchors takes days. 5Profiler’s foundry drafts them from the role’s real work, gates every item twice, and hands your team the pen for final approval.
Before any human time
2 gates
1 · 3 · 5 behaviourally anchored
5 anchor scale
Always human
1 approver
Draft → gate → gate → approve
1 · Ground
The draft starts from the occupation’s real tasks, measured behavioural demands and your declared values.
2 · Validate
A mechanical validator enforces scoring discipline — anchor spread, wording, level separation.
3 · Adversarial review
An independent AI reviewer tries to tear each item apart before a person ever reads it.
4 · Approve
Your team approves item by item; nothing is scoreable until a named human says so.
A 4 should mean the same thing in every room
Unanchored scales measure the rater, not the candidate. Anchors pin the scale to behaviour — “de-escalates live and names the trade-off” — so scores compare across panels, days and interviewers.
- Role-real questions — wrapped in the occupation’s actual tasks — no “greatest weakness” filler.
- Anchored 1-3-5 — each level described in observable behaviour, drafted for your role, edited by you.
- Wording frozen at use — ratings keep the exact question they scored, forever — rubric evolution never rewrites history.
- Bank + variants — approved items join your bank; variant generation keeps repeat candidates from rehearsing.
One role, twelve questions, one afternoon
A Field Sales Manager rubric: the foundry drafted from the role’s eight real task areas and the company’s declared values, both gates passed ten of fourteen drafts, the hiring manager rewrote two anchors and rejected two items — twelve approved questions, live for that evening’s panel. The workshop that used to take two days took a coffee.
Questions we always get
Does AI decide what gets asked?
No — AI drafts and stress-tests; a named human approves every item before it can score a candidate. The audit trail shows who.
Can we write our own items?
Yes — author from scratch or edit drafts; the validator checks your items with the same discipline.
What stops rubric drift?
Versioning with frozen wording: every past rating keeps the exact text it scored against.
Go deeper in the research library: running structured interviews · mechanical vs holistic judgment.
People also evaluate
Forge your first rubric this week.
A 30-minute walkthrough with your roles, not a canned deck — and a sample report to keep.