Structured feedback instead of an opinion

Interview feedback capture works now. An interviewer records feedback against the interview, the feedback list shows every submission against a candidate, and the score dial gives a visual summary of where the panel landed. A recruiter reviewing a shortlist can see who has submitted, who has not, and how the scores compare.

This is real, usable structure and it is already better than a reply-all thread. What it does not yet have is full template configurability, which is the part being extended.

The work in progress is scorecard templates: defining a criterion set per requisition or per role type, assigning criteria to particular stages, and setting a rating scale with written anchors describing what each point on the scale means.

Anchors are the part that does the work. A one-to-five scale with no anchors produces a distribution centred on four, because interviewers are reluctant to write down a two about somebody they have just met. A scale where three is defined as meets the requirement as written and four as exceeds it with evidence produces ratings that mean something. Until templates land, criteria are captured against the interview rather than configured centrally.

A rating on its own is a number somebody chose. A rating with a supporting comment is a claim with evidence, and requiring the comment changes what interviewers do: writing down why forces the interviewer to notice whether they actually have a reason.

It also makes the debrief shorter and better. Instead of reconstructing four impressions from memory, the panel reads what each person recorded at the time and spends the meeting on the criteria where they disagreed. Most panel disagreement, once the evidence is visible, turns out to be about coverage rather than judgement.

Timing is part of the structure. Feedback submitted before an interviewer has spoken to the rest of the panel is worth more than feedback written after a corridor conversation, because a panel that talks first converges on whoever spoke first. Outstanding feedback is flagged as an exception with the interview date attached, so a debrief two days later does not quietly become the moment when four people decide together what they each thought separately.

Averaging a panel into one number destroys the most useful information the panel produced. Two fives and two twos average to a three and a half, which describes nobody's view and hides the fact that half the panel would not hire this person.

Panel views show the spread per criterion alongside any aggregate, so a split is visible rather than smoothed away. A candidate with wide variance is a candidate to discuss, not a candidate to rank. The aggregate exists because people want one, and the spread sits next to it so the aggregate is not mistaken for a verdict.

Where a hiring decision is questioned, whether by a candidate, a client or a regulator, the useful artefact is a contemporaneous record showing what was assessed, by whom, against what criteria, with what evidence. Scorecards produce exactly that as a by-product of doing the interview properly.

Surhires stores that record, keeps demographic data separately from it so it cannot influence a decision it is meant to measure, and exports the whole trail per requisition. Whether your process is lawful in a given jurisdiction depends on how you run it. What the product does is make sure the evidence exists.