Forecast accuracy

Forecast accuracy: predicted vs actual demand per run.

Example

FROM forecast_runs
  SHOW predicted_units, actual_units
  GROUP BY product
  ORDER BY predicted_units DESC

Metric	Type	Description
`predicted_units`	Number	Forecast units (predicted velocity × 7) over the graded weeks.
`actual_units`	Number	Units actually sold in the graded weeks.
`forecast_wape`	Percent	Weighted absolute percentage error: Σ\|actual − predicted\| ÷ Σactual; lower is more accurate.
`forecast_bias`	Percent	Σ(predicted − actual) ÷ Σactual; positive = over-forecasting, negative = under.
`weeks_compared`	Number	Number of graded (variant, location, week) predictions in the group.

Dimension	Description
`product`	Product with image, rendered as a single cell (variants roll up).
`sku`	Variant SKU - splits a product-grain report into variant rows.
`variant`	Variant title (e.g. “Small / Blue”), shown as a dash for products without options. Group by product too - on its own, variants sharing a title across different products fall into one row.
`location`	Fulfilment location.
`vendor`	Product vendor.
`product_type`	Shopify product type.
`forecast_model`	The demand model used for the prediction: linear, trend, day-of-week (dow), seasonal, or intermittent.

Filter	Type	Description
`collection`	Number	Restrict to products in the selected collection(s).
`category`	Text	Restrict to products in the selected category (matches sub-categories too).
`sales_channel`	Number	Restrict to products published to the selected sales channel.
`region_catalog`	Number	Restrict to products in the selected region (market) catalog.
`b2b_catalog`	Number	Restrict to products in the selected B2B (market) catalog.
`company_location_catalog`	Number	Restrict to products in the selected company-location catalog.
`vendor`	Text	Product vendor.
`product_type`	Text	Shopify product type.
`sku`	Text	Variant SKU.
`location_name`	Text	Fulfilment location.
`forecast_model`	Text	The demand model used for the prediction (linear, trend, dow).