What it's built on

kudoSnap collects structured peer nominations and reports what they add up to. The claim is that this carries real, decision-relevant signal about how somebody is seen at work, and that aggregating it honestly is worth doing.

That is the whole claim. It is deliberately narrower than it could be, and the sources below are held to the same boundary: each one is listed with the job it does here and the argument it does not support.

Asking every member of a group to choose someone for a stated criterion — who would you want beside you in a crisis — is a sociometric test, formalised by Jacob Moreno in 1934. kudoSnap polls are sociometric choices in form and function, and the org-level vote graph is a sociogram.

This is not an analogy the product resembles. It is the method the product runs, and it inherits a ninety-year record.

Because the useful part is unreachable any other way. The pane of self-knowledge that is known to others and unknown to you — Luft and Ingham named it the blind area in 1955 — cannot be opened by introspection, by definition.

And because peer judgment is the older tradition, not the lightweight alternative to it. The trait structure that psychology now measures with questionnaires was found in peers describing each other: Tupes and Christal's 1961 analyses, replicated by Norman in 1963 in a paper titled, literally, Replicated factor structure in peer nomination personality ratings. Observers came first. Questionnaires came later.

Hollander's 1954 work on buddy ratings found peer nominations in training cohorts predicted later officer success where evaluation from above missed it, and Connelly and Ones (2010) found that what the people around you say predicts workplace behaviour better than what you say about yourself.

Ask the person-perception literature what structure observer judgment has and it answers, with unusual consistency, two dimensions: is this person effective, and is this person good to us. Rosenberg, Nelson and Vivekananthan mapped it in 1968; Fiske, Cuddy and Glick reached it again as warmth and competence in 2007; Abele and Wojciszke as agency and communion. Different names, same pair, across cultures and instruments.

The four pillars are those two axes, each split once. On the work side: did the work get done, or did it get clearer? On the people side: did people move together, or did someone come away better?

The two axes are the evidence. The splits within them are our design — consistent with the science, not derived from it. We would rather say that than imply a pedigree the split doesn't have.

Not squeamishness. Kluger and DeNisi's 1996 meta-analysis of 607 effects found that input directing attention at the self — threat, deficiency, comparison — reliably makes people worse, while input tied to specific observed acts does not.

The other half of the picture is real and worth hearing. Hearing it takes consent, containment and a skilled person in the room, and a Slack bot has none of the three.

A framework that can't be wrong is decoration. Four predictions, written down before the data arrives:

• 1At scale, the structure in the data should separate the two axes — the people-facing pillars from the work-facing ones. If it doesn't, the pillar map is wrong.not yet testable
• 2People should systematically under-predict the people-facing recognition they actually receive. If anyone has a blind spot, the theory says it points there.not yet testable
• 3People who vote for the same person repeatedly should spread across different facets rather than repeating one. In our own beta, 97% of repeat pairs span two or more — which is exactly why no single devoted voter can mint a card alone.confirmed
• 4As the number of distinct people recognising someone grows, their profile shape should stabilise. If it never settles, we are measuring noise.not yet testable

Not a psychometric instrument. We don't call it one. Nominations identify who stands out at something; they cannot order the middle of a group, and we never claim they describe a person rather than how a person is seen.

Not evaluation data. Nothing here is built to reach a review. You can only be chosen in polls you appear in, so a count is not a ranking and cannot be turned into one.

Not a full picture. Silence is never read as its opposite. A facet with nothing in it means nobody has said it yet, not that it isn't true of you.

Every source, what it contributes, and the claim it does not support.

• Cooley (1902)

  The looking-glass self — the thing this product measures. You assemble your self-image out of what you think others see.

  not That reflected appraisal is all the self is.

• Mead (1934)

  The generalized other — why breadth of distinct voters, not depth of votes, is the unit of a reputation.

  not Treating any one workplace as a finished picture of anybody.

• Moreno (1934)

  The sociometric test — the poll's method and its lineage.

  not Moreno's therapeutic program, which we don't practise.

• Luft & Ingham (1955)

  The blind area — the entire educational purpose, moving strengths from unknown-to-you into known-to-you.

  not The window as a validated instrument. It is a heuristic model.

• Hollander (1954)

  Peer nominations carry predictive signal that evaluation from above misses.

  not Using any of this for evaluation. The product forbids it.

• Tupes & Christal (1961); Norman (1963)

  Trait structure was discovered in peer judgment — observers first, questionnaires later.

  not Claiming kudoSnap's facets are the Big Five, or any validated factor structure.

• Kane & Lawler (1978)

  Nomination is the most reliable peer format, and is best at identifying who is high on a dimension — hence peak claims only.

  not Ordering the middle of a group.

• Thorndike (1920)

  Halo — one judge's impression bleeds across unrelated qualities. The first row of our threat model.

  not Despairing of peer data. Halo dilutes as independent judgments aggregate.

• Kenny (1994)

  Splitting a judgment into the perceiver, the target and the relationship — and reputation as the target part.

  not That kudoSnap computes this properly. It approximates the spirit through breadth gates.

• Rosenberg et al. (1968); Fiske et al. (2007); Abele & Wojciszke (2007, 2014)

  The two axes of person perception — the pillar structure.

  not The splits within each axis. Those are ours, by design.

• Connelly & Ones (2010); Oh, Wang & Mount (2011)

  What observers report out-predicts self-report for workplace behaviour.

  not Marketing any of this as a prediction of someone's future at work.

• Kluger & DeNisi (1996)

  607 effects: input aimed at the self degrades; act-specific praise doesn't. The evidence base for positive-only.

  not That all praise works, or that negative input merely needs better framing.

• Roberts et al. (2005)

  The Reflected Best Self — scholarly precedent for structured, positive-only peer reflection as identity development.

  not Their essay method. Ours is nomination, which is a different instrument.

• Galton (1907)

  Independent judgments, aggregated, beat individual ones — the reason for the breadth gate.

  not Reading it as anything more than that.

• Deci & Ryan (2000)

  Why the loop holds attention: it feeds real needs for competence and relatedness.

  not Gamification. Streaks and badges are not this.

• Edmondson (1999)

  The org-level read: broad people-facing recognition as a signal adjacent to psychological safety.

  not Claiming kudoSnap measures psychological safety. It doesn't.