Skip to content
Demand Forecasting

Forecasts you can actually trust.

Ensemble of 5 proven models. Seasonality without trend extrapolation. Realized-APE tracking so you know when the model earns your trust.

app.buycast.ai/overview
AB
Demand forecast with realized-APE telemetry per SKU.
Demand Forecasting in the demand forecasting dashboard.
What it does

Capabilities

Statistical rigor meets production discipline. Every forecast comes with the receipts of how well that same method has performed on that same SKU, historically.

Ensemble of proven models

AutoETS, AutoARIMA, AutoTheta, AutoCES, and MSTL on every plan. CrostonOptimized for intermittent SKUs.

Change-point detection

Recent regime shifts dominate fitting — but full history still drives seasonal profile. Your launches, stockouts, and price changes no longer poison the forecast.

Daily shadow forecasting (Pro)

A 20% sample runs every candidate method every day so method A/B testing is continuous, not a one-time bake-off. Available on the Pro tier and above.

Realized-APE telemetry

`forecast_actuals` continuously tracks realized accuracy per SKU per horizon. Your purchasing agent auto-downweights untrustworthy forecasts.

How it's different

Built for operators who've outgrown spreadsheets.

Not a black-box ML model

Classical statistical methods for 95% of your SKUs — reproducible, auditable, explainable. Neural models only where they measurably win.

Accuracy you can verify

Every forecast is scored against realized demand. You see, per SKU, whether the model has earned your trust.

Under the hood

Show your homework.

Every claim in this product is instrumented. Here's how.

  • Runs on a dedicated batch worker with parallelized SKU processing
  • Method selection per SKU based on rolling realized APE
  • Seasonality profile derived from full multi-year history; trend from recent regime
  • Integrated with buy-box share and stockout history for demand correction

Stop reacting. Start compounding.

14 days free. Your data stays yours.