SETPOINTk.ai · Metric Surface

SPI · Method

SPI is computed from longitudinal features over rolling windows. This page describes the *shape* of the method without exposing proprietary weighting/transform details.

1) Inputs

Public-safe examples (not exhaustive)

HRV-derived signals (rest/recovery dynamics)
Sleep timing + continuity
Activity + load proxies
Self-reported recovery + context (optional)

2) Windowing

Rolling feature windows

Compute features on multiple window lengths (e.g., short/medium/long)
Emphasize longitudinal behavior over isolated readings
Track regime stability and transitions

3) Feature families

What gets summarized

Stability: baseline coherence + bounded variability
Recovery: post-perturbation slope + consistency
Variability structure: organized vs noisy variability
Drift: sustained baseline movement across time

4) Composition

Composite surface, not a single test

SPI is a normalized composite of longitudinal feature families. Internal transforms and weights are intentionally not disclosed on the public surface.

Inputs(t) -> windowed features -> normalized families -> composite surface
SPI(t) = f( stability(t), recovery(t), structure(t), drift(t) )  // public-safe form

5) Interpretation

Zones are a grammar

SPI supports a zone grammar: stable / recovering / volatile / drifting
Zones describe *behavior states*, not diagnoses
Clinical usage (future) is authenticated, governed, auditable

Boundary

What SPI method is not

Not a diagnostic test
Not a triage engine
Not a treatment recommendation
Not a replacement for clinician judgment

Public-safe disclosure

This surface is educational and interpretive. It intentionally avoids exposing internal transforms, thresholds, or weighting. Any future clinical-mode implementation operates under separate posture, policy, access controls, and audit logging.