Upload consumption history by SKU. Every SKU gets its own model competition across thirteen time series models, scored on data held back from fitting, and a 24-month forecast from whichever model earned it.
Your data has been loaded, please map the columns below so the tool knows which column is which.
Three columns, one row per SKU per month, stacked — not a column per month.
| SKU | Period | Quantity |
|---|---|---|
| A-1001 | 2024-01 | 1,250 |
| A-1001 | 2024-02 | 1,310 |
| A-1001 | 2024-03 | 1,180 |
| B-2044 | 2024-01 | 88 |
| B-2044 | 2024-02 | 94 |
| B-2044 | 2024-03 | 0 |
Headers can be named anything — you map them in the next step. Periods accept 2024-01, Jan-24, 01/15/2024 and similar. Zero is a real value and should be recorded; leave a month out entirely and it gets filled as zero demand. Add as many SKUs as you like — each one gets its own model competition.
| SKU | Periods | Winning model | MAPE | WAPE | Bias | Next 24m | Status |
|---|
One row per SKU, one column per forecast month, from that SKU's own winning model. Scroll sideways for later months.
| Model | MAPE | WAPE | MAE | RMSE | Bias |
|---|
Ranked by MAPE. WAPE weights errors by volume and survives zeros — if the two disagree, trust WAPE. Bias is signed: positive means the model over-forecasts.