QR Demand Planning
Sign In

One forecasting engine. Three ways to use it.

QR Demand Planning is one forecasting engine. You can use it as a web app, as a model alone connected to your infrastructure, or as an app fed directly by your sales data — on our cloud or yours, under the same governance.

One forecasting engine with three inbound paths and three destinations. Press 1, 2 or 3, or use the arrow keys, to inspect a path.

Inbound

Spreadsheet or CSV

uploaded by the planner

Forecasting engine

Outbound

Web app

configure · run · analyze

1 · The planner uploads a spreadsheet; the web app helps configure the run, executes the engine and shows the result.

Where the accuracy gain comes from

AI at its limit powers advanced forecasting.

Analytical engine

  • Library of machine learning and statistical models on scalable infrastructure, testing hundreds of hyperparameter combinations.

  • Models run for every Demand Forecast Unit (DFU), with the ideal model combination recommended for maximum accuracy.

  • Backtesting of past algorithm performance over the planned horizon.

  • Plus outlier detection, demand normalization for working days and holidays, ABC classification, hierarchical reconciliation, external regressors such as price or promotions, quality scores per series and prediction ranges.

AI at full potential to raise accuracy

Hundreds of combinations tested (illustrative)best combination per DFU

Hyperparameter combinations are tested across algorithm classes and combined per DFU for more robust forecasts.

What comes out is a forecast with confidence bands, the scoreboard of every model tried, and the backtest that justifies the winner. Everything below is that same engine. Only the way you reach it changes.

Beyond the standard forecast.

Two situations where a product's own history is not enough, and how the engine handles them.

Cold start

New or short-history series: the forecast comes from analogous series rather than their own past. In Results this is labeled New product.

Coming soon

Cannibalization

When a new item or a promotion takes sales from products you already sell, their history alone overstates future demand. Cannibalization modelling estimates that transfer, so the forecast for the whole portfolio adds up. On our roadmap.

Choosing between the three.

Three questions usually settle it: where your sales history lives today, who runs the planning cycle, and how often the forecast needs to be refreshed. The engine is the same in every case, so starting with one option and moving to another later does not mean starting over.