When you post a role, the assessment selector picks the right mix from the library based on the role family, the must-haves, and the seniority. An SDR does not get a numerical reasoning battery they will never use. A senior engineer does not get a coding task that screens out by language preference. The selection is visible to you and editable before the role goes live.

• ●Role family maps to a default battery (e.g. AE → personality + SJT + 20-minute role-play).
• ●Must-haves modify it (e.g. "must have managed a quota team" adds a leadership SJT module).
• ●Seniority modifies length (junior 25 min, senior 40 min, head-of 60 min).
• ●You can swap any module for an alternative before the role publishes.

On cognitive batteries, the next item depends on the last response. A strong run pushes the candidate into harder items quickly; a weak run keeps them at a fair level. Two outcomes: a more precise score, and a shorter test for the candidate. The median assessment runs 22 minutes, not 45.

Every assessment in production is monitored for adverse impact against protected groups, per role, per stage, per audit cycle. The 4/5ths rule is the floor. When a flag fires, the role's hiring manager and our internal review team both see it before the next candidate is processed.

• Annual independent bias audit, summary published.
• Per-role drift detection: if a role's pass rates shift on a protected attribute, we know and so do you.
• Adverse-impact dashboard in product, shareable with HR or legal.
• See /trust/fairness for the full posture.