Scientific methodology

What does Tanbu measure?

Dimensions

Tanbu shows organizational health across six dimensions (engagement, leadership, communication, development, well-being, and culture) on a 0‑100 scale. Dimensions that clearly stand out from the rest are flagged only after a statistical test, and scores for groups with too few responses are not shown.

Framework

Questions are tied to a measurement framework of dimensions, constructs, and sub-constructs.

The constructs draw on established academic frameworks.

Scientific panel

The scientific panel on the survey screen explains in plain language which research area each measured dimension draws on and whether consistency can be measured yet.

How is a score calculated?

The dashboard and analytics screens explain how scores are calculated: reverse-coded items are corrected, scores are put on a 0‑100 scale, every respondent counts equally and a confidence interval is shown around scores.

Minimum responses and uncertainty

Scores, reliability values, and relationship analyses require a minimum number of responses. Below the threshold the result is withheld or labeled preliminary.

Scores are aggregated at respondent level and, with enough responses, each dimension score is shown with a 95% confidence interval.

What does Cronbach's alpha show?

Alpha shows internal consistency. It is not evidence of validity.

Internal consistency

For surveys with at least three scored items in a dimension and enough complete responses, internal consistency is measured with Cronbach's alpha. Until those conditions are met, no reliability value is shown.

Item analysis

Once enough responses are in, each item's fit with its scale and its discrimination are computed, and the survey panel separates solid items from items worth reviewing.

eNPS

Tanbu calculates eNPS, which measures how likely employees are to recommend the organization, at survey and company level, together with the shares of promoters, passives, and detractors. It withholds the score when there are not enough responses.

Relationships and risk level

Relationships

Relationships between dimensions are shown only when they pass a significance threshold, and the screen states plainly that they are associations, not causes.

Risk level

Tanbu derives a low, medium, or high risk level per dimension from survey scores and shows none when there are not enough responses.

It is not an attrition prediction.

Change over time

Tanbu shows how scores change across monthly, quarterly, or yearly periods. Periods with no data, or withheld for privacy, stay empty, and the line is not joined across them.

AI commentary

AI-written narratives are kept within what the data supports: a weakly evidenced finding is not stated in definitive language, and interpretation is withheld when data is insufficient.

Tanbu's AI layer, Helia, is designed to interpret group-level findings and does not make decisions about individuals.

Small groups and anonymity

Small groups

In reports and analytics views, scores are not shown for groups too small to protect the people in them. A group's score appears only once enough people have responded.

Anonymous surveys

Targeted anonymous surveys cannot be sent to very small audiences, and the anonymity setting cannot be changed once a survey is deployed.

In an anonymous survey, the response is stored without the respondent's name or account.

Method in reports

Executive reports include a trust and methodology page covering sample sufficiency, the anonymity floor, how to read scores, the data period, and an identifier of the data snapshot behind the report.

Source information

Figures on the dashboard and team report are shown with their source and scope.

What Tanbu does not do

• Tanbu does not administer published instruments as they stand.
• Alpha shows internal consistency. It is not evidence of validity.
• No reliability value is shown for short pulse surveys.