In development
An evaluation-first engine: data sufficiency gates, chronological backtests and calibration before any forecast is shown.
Not validated. Gates and backtests available; forecasts gated.
What you see today
Illustrative example — not a client result
For the operator choosing the next improvement with more context—and less guesswork
Explore an explicit simulation now, while the real forecast remains unavailable until first-party data passes validation gates.
See the workflow
Use the controls to move through a deterministic sample. No customer result or live provider data is implied.
Change assumptions
Two different questions
The engine defines separate targets and eligible data. Sessions and orders belong to product, channel and time; AI observations keep their own denominator.
Validation first
Chronological backtests, product-group isolation and leakage checks run before an output may be called a forecast.
Uncertainty visible
The readiness view shows missing days, groups or sessions. It never fills the gap with a simulated client result.
What you take away
Clear data requirements and exclusions.
Error and calibration metrics when enough eligible history exists.
Action snapshots and outcomes for future evaluation—without causal claims.
Before you start
No. The v0.1 framework and gates exist, but Gengine does not advertise a validated predictive model.
No. Training is workspace-only unless a future explicit opt-in says otherwise.
Not validated. Gates and backtests available; forecasts gated.