How a KEII® score is calculated

Every assessment is scored against the same six-dimension framework. The models score each dimension; Kloutify then combines the six dimension scores into the overall index on its own servers. The overall index is a proprietary weighted aggregate of the six dimension scores. The weights are fixed, applied identically to every profile, versioned with the methodology, and available to auditors under NDA.

A live assessment prompts up to three independent large language models over the same profile and the same framework. Each model returns an overall score and six dimension scores. Kloutify then applies a deterministic aggregation: confidence-weighted averaging per dimension, cross-model agreement and outlier detection, reliability grading, and clamping into the 0 to 100 range. The aggregation code is covered by deterministic unit tests; the individual model judgements are, by nature, the models’ own.

Model output is schema-validated before it is used. A malformed or non-conforming response causes that model to fail closed rather than contribute a partial number. Profile fields interpolated into prompts are sanitised against prompt injection.

When no model is available, the platform falls back to a deterministic weighted-factor calculation over the supplied profile, content and network inputs. That fallback is labelled as such in the response and contains no random component.

Every stored score carries its scoring method, the per-model breakdown where models were used, and a calculation method label. An integrity guard in the production configuration refuses to emit an unverified score: when demo mode is off, the platform serves only model-computed or deterministic values and returns an honest “unavailable” state instead of a placeholder.

The free public assessment on /score is a twelve-question self-assessment scored server-side by a fixed formula. It is labelled as a self-reported baseline and is carried into the account when the visitor signs up, so the number does not change between the public page and the dashboard.

KEII is not a patented method and it has not yet been validated against an external ground-truth corpus; the weighting is expert-defined and is described as such. It is a diagnostic indicator, not a ranking, a credit score, a social score or an employee monitoring signal. Use of the platform for employment decisions is expressly prohibited under the terms of use.

The remaining work to graduate the engine to fully validated status is public: a calibration corpus, an adversarial test suite run with live providers, and a full input snapshot on every stored result. Those items are tracked in the methodology of record and will be reflected here when they close.