Noetra reads your interview notes, asks the candidate a short structured assessment, and tells the next interviewer exactly what to ask. It never tells you who to hire.
10 candidates free. No card needed.
I paused the cutover and rang the data owner before anything else.
“Walk me through the first decision you made that week.”
Interview panels rarely share a record. Each interviewer forms an impression, writes a few lines, and the next one starts from nothing. The quiet candidate and the practised candidate get read on practice.
Three interviewers, three private readings, and nothing that can be checked later or explained to a candidate who asks.
Without a shared record the second round re-covers the first, instead of going deeper where it would matter.
Practised answers are easy to read and easy to rehearse. What someone did under pressure is harder to ask about well.
Set up a role once. After that it is paste the notes, read the guide before the next round.
Paste the job description. Noetra reads out the requirements and you switch off anything that does not apply. Four short questions about what the role actually demands set the emphasis for every guide.
About two minutes, once per role.
Upload a resume if you have one: names and contact details are removed before anything reads it. After the round, paste your notes. The candidate is asked to consent, then takes a short structured assessment.
One credit, used when the assessment is sent.
What the evidence shows on each dimension, how confident that reading is, and the specific questions worth asking next. After each round the guide updates with what was asked and what came back.
Every round included, no extra charge.
The same five for every candidate and every role. What changes between roles is which ones the guide leans on, never the questions and never the scoring.
Planning, follow-through, and the care taken when nobody is checking.
How the work gets done with other people: handovers, disagreement, who gets the credit.
What holds when the week goes wrong: pace, priorities, keeping people informed.
Where the line sits on shortcuts, and how those choices get explained afterwards.
Deciding on incomplete information, and knowing the moment to escalate.
A percentile for this role family. Never a grade, and never a ranking against your other candidates.
How much the reading can bear. One weak source caps it, so a thin picture can never look certain.
The candidate words, the items and the references the score came from, traceable one by one.
Specific to this candidate, this evidence and this round, and tracked so it is not asked twice.
The assessment, the interview notes and the references point the same way.
The sources disagree. The guide gives you the question that would settle it.
Not enough has been seen yet. The dimension stays open rather than guessed at.
These are not settings. They are built into the data model and enforced by tests that fail the build, because a hiring tool that can quietly turn into a verdict machine eventually does.
There is no combined rating anywhere in the product, the interface or the data. Five dimensions, read separately.
Noetra never closes a candidate for you. Moving someone forward, or not, is a human action and is recorded as one.
Interview material enters as text. Nothing is read from faces or recordings, and nothing is inferred from how someone sounds.
Nothing is pulled from social profiles or from any data the candidate did not knowingly provide.
A score always carries how much to trust it. Where the evidence is thin, Noetra says so instead of filling the gap.
One assessment, the same for everyone applying for the role, about 30 minutes in four short parts. Built on published work about structured hiring rather than invented for the occasion.
Scoring is deterministic code, not a model forming a view. Models are used for extraction and phrasing only: reading a resume into a timeline, pulling the verbatim line that supports a score, wording the next question. Each of those is checked against the source text before it reaches you. Items are drafted and under psychometric review, and norms stay marked provisional until your own data replaces them.
If a candidate asks what was held against them, you should be able to answer. That constraint shaped the whole product.
Names, contact details and dates of birth are stripped from a resume before any model or person reads it.
Every candidate record carries a purge date, and the job that enforces it runs every hour.
Who sent what, when a score changed, which questions were asked in which round.
No assessment is sent until the candidate has agreed, and they can see what is being shared.
No, and it is built so that it cannot. It produces a score per dimension, the evidence behind it, and a question to ask. There is no ranking of people against each other, no single number, and no action in the product that closes a candidate automatically.
They consent, then work through four short parts: a work style inventory, a work standards scale, a short terms check, and workplace scenarios. About 30 minutes, it saves as you go, and it works on a phone. Naming referees is optional.
Three ways. Scoring is deterministic code rather than a model forming an opinion, so the same answers always produce the same result. Every score carries a confidence value. And the output is evidence plus a question for the next interview, never a label on a person.
No. Paste or upload your interview notes and everything works. Evidence from an interview is deliberately weighted low in any case, because what someone says in a room is the weakest signal in the set.
A fairness suite runs on every release and the build fails if impact ratios fall below threshold. Both scores and confidence are tested, and the reports sit in your workspace. This is our own testing rather than legal advice, and it does not replace your own validation.
Provisional norms per role family at the start, clearly marked as provisional. Once 80 of your own candidates have completed an assessment for a role family, your workspace can be scored against your own population instead.
Set up a role in two minutes and run your first candidate today. 10 are free, and nothing is charged until you pass them.