Nothing is a black box
Every recommendation exposes the evidence rows, the statistical test used, the rules that fired, the alternatives that were rejected and a timestamped audit trail.
Point it at a dataset and it profiles the schema, infers the business domain, derives the KPIs that matter, tests every candidate driver for statistical significance, and writes the executive brief. No dashboard authoring, no configuration, no backend.
Your data never leaves your device. All analysis runs locally in your browser.
CSV, Excel, JSON or Parquet parsed in your browser with DuckDB-WASM.
The domain is inferred, then the KPIs that matter for it are derived.
Every driver is tested for significance and ranked by contribution.
Deterministic rules produce owned, auditable actions.
A written executive brief you could send without editing.
Each one carries a real analytical story: a planted regression that the engine has to find on its own. Pick a domain to open the workspace.
Every recommendation exposes the evidence rows, the statistical test used, the rules that fired, the alternatives that were rejected and a timestamped audit trail.
Drivers are ranked by contribution and screened with Poisson, chi-square, Mann-Whitney and Mann-Kendall tests, with Benjamini-Hochberg correction across the candidate set.
Datasets are scored for quality, freshness and trust, then assigned a decision-readiness level. Poor data caps the confidence of every recommendation derived from it.
Sensitive columns are detected by name and by value, masked automatically, and excluded from drill-down until you explicitly allow it.
DuckDB-WASM registers your file as a table. The analytics run as SQL you can read, and the SQL console lets you query the same tables directly.
The analytics core is a reusable Python package with per-domain plugins that supply KPIs, dimensions, root-cause rules and forecast settings.