Composite scoring and blind ranking
A ranking is a reading order, not a verdict. This page explains exactly what the number is made of, who controls it, and the one thing it is structurally forbidden from doing.
Composite scoring combines a candidate’s verified assessment result, interview performance and role-fit signal into a single 0 to 100 score used to order a pipeline. On CalHire the three weights are set by the employer and must total 100; the default split is 40 percent test, 35 percent interview and 25 percent role fit. The score ranks candidates for human review. No score threshold rejects anyone automatically.
Last reviewed
The short version
- Three inputs, three weights, always summing to 100. You choose the split; the default is 40/35/25.
- The score orders a list for a human to read. It cannot advance, reject or filter anyone out on its own.
- Candidates below your threshold are flagged for review, not discarded, which is a deliberate architectural constraint rather than a setting.
- Every score, weight and human decision is written to a hash-chained audit trail you can export.
What goes into the number
Three components, each measuring something different, each visible on its own.
Test — 40 by default
The verified skills assessment result, weighted across the skills your role requires. This is the component with evidence behind it, which is why it carries the largest default weight.
Interview — 35 by default
Performance in the structured text interview, scored against a fixed rubric rather than an impression. Structure is what makes an interview score comparable between two candidates.
Role fit — 25 by default
How well the candidate’s verified skill set overlaps what the role asked for. A strong generalist and a precise specialist can land in different places here, which is usually the point.
Why the weights are yours and not ours
A default weighting is a claim about what matters, and no vendor is in a position to make that claim about your role. A support role that lives in written conversation should weight the interview heavily. A data engineering role probably should not. A team hiring for an unfamiliar skill set may want role fit close to zero, because they do not yet trust their own definition of fit.
So the weights are set per employer and per role, they must total exactly 100, and the split you used is stored on the decision record. That last part is the one people overlook. A score with no record of how it was weighted is not auditable, and an unauditable score is exactly the kind of thing a regulator asks about first.
The 40/35/25 default exists because a blank form is worse than a defensible starting point. It leans toward the component with the most evidence behind it and away from the one most exposed to interpretation. Change it the moment you have a reason to.
The rule that no score can break
No path in the platform auto-rejects a candidate. Not a low composite, not a failed threshold, not an integrity flag. A candidate below your bar is surfaced to a human with the reason attached, and a person decides.
This is a constraint in the system rather than a configuration option, and it is worth being clear about why. An automated rejection at scale is the specific thing regulators have moved against. The EU AI Act treats employment decision systems as high risk and requires meaningful human oversight. New York City requires an annual bias audit for automated employment decision tools and disclosure to candidates. Colorado requires notification when algorithmic discrimination is discovered. All of that becomes vastly harder to satisfy if the machine is allowed to close a door by itself.
There is also a plainer reason. Scores have error bars. Treating a number as a verdict pretends otherwise, and the cost of that pretence falls entirely on the person who was never read.
What a reviewer sees
- 1
A ranked, anonymous list
Candidates in composite order, each with the three component scores broken out. If someone is high on test and low on interview, that is visible rather than averaged away.
- 2
The reason for the rank
Which skills scored where, and against which weights. A ranking you cannot explain to the candidate is one you should not be acting on.
- 3
Flags, not filters
Below-threshold candidates and integrity signals appear as flags in the same list. Nothing is hidden from you by default, because a filter you forgot you set is a decision you did not know you made.
- 4
Your decision, recorded
Advancing or declining is an explicit action attributed to a person. The score, the weights, the flags and the human who acted all land on the same audit record.
What this does not do
A composite score is not a prediction of job performance and is not presented as one. It is a defensible ordering of the evidence you chose to collect. Anyone selling a hiring score as a performance forecast is describing a much harder problem than the one they solved.
Ranking cannot correct a badly specified role. If the required skills are wrong, the ranking will be confidently, consistently wrong, and it will look rigorous while doing it. Garbage in has never been more legible than when it arrives sorted.
The three components are not independent. A candidate who tests well often interviews well, so a high composite can reflect one underlying ability counted twice. Reading the components separately, which the interface encourages, is a partial defence rather than a complete one.
Scoring does not remove your obligation to check outcomes. Any consistent ranking can produce adverse impact. Selection rates by group need measuring against the four-fifths rule regardless of how principled the weights felt when you set them.
Questions people actually ask
Can we set a minimum score that auto-rejects candidates?
Do the weights have to add up to 100?
Can we change weights partway through a live role?
Does the score include anything about the candidate’s background?
Can a candidate challenge their score?
What happens to a candidate flagged by the integrity layer?
Related
Where this connects to the rest of the platform.
Verified skills assessment
One supervised assessment produces a skills profile an employer can check, instead of a resume they have to take on trust.
AI governance, bias auditing and the audit trail
Adverse-impact analysis on the four-fifths rule, a hash-chained audit trail, candidate appeals, and DSAR handling. The evidence exists before anyone asks for it.
Zero ghosting and candidate report cards
Every candidate who does not get the job receives a report card explaining where they stood and what to work on. It is automatic, and it is not optional.
Read the reasoning
The evidence and the argument behind what is on this page.
Structured interviews: the highest-return change most teams never make
Same questions, same order, same rubric, scored independently. Structured interviews are unglamorous, uncomfortable to adopt, and the best value in hiring.
ReadSee a verified pipeline for one of your roles
Post a role free and review anonymous, skill-ranked candidates. No card, no sales call to get started.