Hire or no hire: a decision framework you write before the first interview

A hire or no-hire decision framework is a short written rule, agreed before the first interview, that turns a panel's scores into a verdict. It names the bar, lists the gates (requirements where one evidenced miss means no), says how the remaining scores combine, and says what happens when the evidence is thin. Interviewers score alone; then one owner applies the rule. The point is that the outcome depends on the evidence and the rule, not on who spoke first or most senior.

This page covers the rule itself: how to write it, a worked example, and what to do with thin evidence and split panels. The meeting where it gets applied is in the interview debrief template, and choosing between several candidates who all clear the bar is in how to compare candidates after interviews.

What goes into a hire or no-hire framework?

Five parts, all written down before the loop starts:

1. The bar. What "hire" means for this role, in one or two sentences.
2. Gates. Two or three requirements the role cannot do without. A clear, evidenced miss on any gate is a no, whatever the other scores say.
3. Tradeable criteria. Everything else on the scorecard. Strength in one can offset a gap in another.
4. The combination rule. How scores become a verdict: for example, "every gate at 3 or above on a 1 to 4 scale, and no tradeable criterion at 1."
5. The evidence rule. What counts as enough evidence, and what happens when a gate was not properly tested.

Plus one name: the person who applies the rule and makes the call, usually the hiring manager.

Why write the rule before the interviews?

Because people combine evidence worse in their heads than a simple rule does. A meta-analysis by Kuncel, Klieger, Connelly and Ones (2013, Journal of Applied Psychology) compared mechanical combination of selection data with holistic judgment. It found a consistent and substantial loss of validity when data were combined holistically, even by experts who knew the jobs; for predicting job performance, the mechanical methods improved prediction by more than 50%. Interview research points the same way: in the meta-analysis by Conway, Jako and Goodman (1995), multiple ratings improved reliability when they were combined mechanically, with no evidence of a benefit when they were combined subjectively.

There is also a timing reason. A rule written after the interviews tends to be written around the candidate the team already likes. Highhouse (2008) describes the belief that experience makes intuition reliable as one of the main reasons selection aids get ignored. Writing the rule first is the cheapest guard against that.

How do you define the bar?

There are three common ways, and it helps to say out loud which one you are using:

For a hire or no-hire call on one candidate, use an absolute bar. The comparative question comes later, once you know who cleared it.

What does a written decision rule look like?

Copy this and fill it in at the intake meeting, before any interview is booked:

HIRE / NO-HIRE RULE
Role: [title]            Version: [v1]        Agreed on: [date]
Decision owner: [name]   Agreed by: [hiring manager, recruiter, panel lead]

THE BAR
"[One or two sentences: what a hire can do in the first six months.]"

GATES (one evidenced score below 3 = no hire)
1. [competency] - tested in [interview], by [interviewer]
2. [competency] - tested in [interview], by [interviewer]
3. [competency] - tested in [interview], by [interviewer]

TRADEABLE CRITERIA
[competency], [competency], [competency]
Rule: no score of 1; at most one score of 2

SCALE (1-4): 1 = clear gap, 2 = below the bar, 3 = meets the bar, 4 = above

EVIDENCE RULE
A gate score needs at least one example in the candidate's own words.
Gate not tested, or tested without evidence = "insufficient", not a 2.
Insufficient on a gate = one targeted follow-up, then decide.

VERDICTS
Hire | No hire | One follow-up (named question, owner, date)

CHANGING THE RULE
Only between candidates, never during a debrief, and the new
version applies to everyone still in process.

The rule then reduces to a small decision table:

A worked example: three candidates, one rule

An example, with invented candidates and scores. The role is a customer success manager. The gates are account retention, handling escalations and written communication. The tradeable criteria are product knowledge, forecasting and coaching others.

• Avery: hire. Every gate is at 3 or above and there is one 2 on a tradeable criterion, which the rule allows.
• Blake: no hire, despite the highest total. The 2 on escalations came with evidence: asked twice for an example of an angry customer, Blake described handing the call to a manager both times. A total would have put Blake first and hidden the one gap the team said it could not accept.
• Casey: one follow-up. The escalation question was skipped when the interview ran long. That is a process gap, not evidence of a weak candidate. The owner books a 20-minute follow-up with one escalation question and a probe, and decides within two days.

The arithmetic: Avery 3 + 3 + 4 + 2 + 3 + 3 = 18; Blake 4 + 2 + 4 + 4 + 4 + 3 = 21. Notice that the rule never uses the totals. If you want a weighted total for the tradeable criteria, the weighted interview scorecard shows how, but apply the gates first.

What should you do when the evidence is thin?

Mark it as insufficient, not as a low score. A missing example is a fact about the interview; a low score is a claim about the candidate. Then cap the fix at one step:

• Name the gap: which gate, and what evidence would settle it.
• Choose the cheapest step that tests it: a short follow-up call, a work sample or a targeted reference question. Not a whole extra round; see how many interview rounds is normal for why extra rounds add less than they cost.
• Set an owner and a date. "Let's think about it" is not a verdict.
• Fix the loop for the next candidate, so the gate is tested on time.

How should a split panel be resolved?

The framework only works if each interviewer's score is their own. Scores are submitted before anyone discusses the candidate; independent interview feedback before the debrief explains why and how to enforce it. In the meeting, compare evidence on the gates first, then the tradeable criteria. A score changes only when someone names evidence the interviewer did not have.

If the owner overrides the rule, that is allowed; it is their decision. What is not allowed is overriding it silently. The record should say which evidence outweighed which and what risk is being accepted. Some teams add an independent reviewer from outside the hiring team for this reason; the bar raiser interview page covers how Amazon does it.

What should be written down after the decision?

• The verdict, the date and the owner.
• The rule version used.
• Each gate's score with the candidate's own words behind it.
• Any score changed in the debrief, with the new evidence.
• Any override, with the reason and the accepted risk.
• For a no, the job-related reason, in a form you could share with the candidate if asked.

The debrief template has a decision record to copy. If the evidence lives in an AI scorecard, check that every quote really appears in the transcript; in Interview Signal, for example, a quote that does not appear in the transcript is removed from the scorecard.

How do hire or no-hire frameworks break?

• Gates rewritten mid-search to fit a candidate someone likes. Change the rule only between candidates and apply the new version to everyone still in process.
• Averaging across gates, so a 4 on one gate covers a 2 on another. Gates do not trade.
• "Culture fit" as a gate with no anchor. If nobody can say what a 3 looks like, it is not a gate; see how to assess values without culture-fit bias.
• Deadline pressure. OPM's structured interview guide lists "pressure to hire" among common interviewing mistakes: decisions made on a limited sample of information. An open seat is not evidence.
• Too many gates. Six gates means almost nobody passes and the team quietly stops using the rule. Two or three is enough.