PhasePoint discovers latent temporal zones directly from your timestamps and exports two bundles: model-ready features and human-readable evidence. You uncover time structure without manual feature engineering, and receive reports (zones, support/lift, stability) that make time effects reviewable and easy to integrate into any workflow.
Find regimes
Improve forecast
Explain effects
Review results
Feature export
Model agnostic
On-prem ready
Diagnostics & evidence
Find regimes
Improve forecast
Explain effects
Review results
Feature export
Model agnostic
On-prem ready
Diagnostics & evidence
What PhasePoint is not
Not a new model family: PhasePoint exports features and evidence; your forecasting/ML model stays the same.
Not a new model family: PhasePoint exports features and evidence; your forecasting/ML model stays the same.
Not a new model family: PhasePoint exports features and evidence; your forecasting/ML model stays the same.
Not opaque automation: you get a zone catalog + stability diagnostics, not an unexplainable transformation.
Not opaque automation: you get a zone catalog + stability diagnostics, not an unexplainable transformation.
Not opaque automation: you get a zone catalog + stability diagnostics, not an unexplainable transformation.
Not calendar engineering: beyond day/month/holiday flags, PhasePoint finds latent temporal regimes from the data itself.
Not calendar engineering: beyond day/month/holiday flags, PhasePoint finds latent temporal regimes from the data itself.
Not calendar engineering: beyond day/month/holiday flags, PhasePoint finds latent temporal regimes from the data itself.
Not a new model
Not a new model
Not basic flags
Not basic flags
No black box
No black box
How it works
Ingest:
Timestamp + Target (or residuals)
Discover:
Zones with support/lift + stability checks
Generate:
Model-agnostic feature matrix
Export:
Feature bundle + evidence bundle, ready for downstream models
Outputs: Feature, Evidence Bundles
Feature Bundle
Format: CSV / Parquet Use: join to training data → train any model
Exported time features you can plug directly into your pipeline without refactoring your models.
Zone membership and proximity
Time-geometry features
Optional stability-weighted variants
Zone membership and proximity
Time-geometry features
Optional stability-weighted variants
Format: CSV / Parquet
Use: join to training data → train any model
Evidence Bundle
Evidence Bundle
Purpose: transparency, adoption confidence, and audit-ready review.
A structured report bundle that makes PhasePoint’s outputs easy to trust, understand, and review.
Zone catalog: definitions, names, intervals, and metadata
Detect missed time effects. Reveal recurring, interpretable patterns that common seasonality features and standard models often miss. Especially helpful in multi-scale data.
Lift performance without refactoring. Drop in new features; keep your model family.
Explain and validate the gains. Evidence, stability checks, and reports come with every run.
