# The sell-out forecast

Source: https://docs.tightly.io/help/connect/forecast-engine-overview
Reviewed: 2026-10-29

## The question this page answers

What does the engine expect this account's doors to sell over the coming weeks, and how sure is it?

## What you see


**Account order plan**: **Save page** → **Review assortment** → **Save draft** → **Approve assortment**. Save **Dispatch dates** and warehouse.

**Review channel inputs** checks all Accounts and offers windows. **Review consumer baselines** previews a baseline for each variant; **Use baseline** saves a review.

**Review forecast preparation**, then **Request first forecast** where enabled. **Previous requests** finds earlier requests. **Check progress** reads status; **Check inputs** compares original inputs. Saving plans starts no forecast or order.

Open an account's **Forecast** view for the latest engine run, summary figures, weekly chart and grid. **Product plans** below the grid holds saved plans.

- **Commitment, units** shows the P50 total for exact Commitment dates. Partial data is marked **Forecast subtotal, units**. Rolling views show **Next 4 weeks, units**, **Next 13 weeks, units** or **Next 26 weeks, units**. The grid uses P50; **P10 to P90** appears only when both bounds are supplied.
- **Accuracy, latest forecast** shows **WMAPE** once scored. Hover or focus it for **Bias** in words and **N of M doors scored**. Until then it reads a dash, with the reason in the same tip. [Sell-out forecast accuracy](../connect/sell-out-engine-accuracy) explains both.
- **Doors measured** counts doors with their own reported sales. The rest take a share; **Basis** says whose. The **N not scored** chip opens the doors the last forecast could not be scored on, each with its reason and basis.

Focus the run chip for timing and edited-cell count.

One row of controls follows: **Styles**, **Categories** and **Doors**, then 4, 13 or 26 weeks and **Commitment** where a Commitment is selected. Commitment opens on its exact dates and keeps only stored weeks that overlap that window. **Partial forecast** or **No forecast** identifies missing coverage. On **Doors**, use the visible **Product** selector, **Find a door**, and **Filters** for store group and **Edited doors**. The **...** menu holds **Shipment schedule** (only when a Commitment is selected), **Reorders** and the not-scored doors. Shipment schedule keeps the selected product; **All styles** opens account scope. Saved pending requests retain their original scope.

Categories sum products and retain ranges only with complete coverage. **Doors** sums **All styles** or the selected **Product**. Rows show store group, size grade and **Basis**:

| Basis | What stands behind the door's number |
|---|---|
| Measured | The door's own reported sales |
| Representative | A sister door's sales |
| Store group | Doors of the same type |
| Category | The category's average |
| Account average | The account's average door |
| Predecessor, Kit, Look-alikes, Family, Sell-in | The product's baseline (product rows) |
| Baseline | Its own weeks |

The chart spans the workspace independently of the grid. It draws the forecast as a dashed line and shades only supplied P10 to P90 bounds. Reported weeks use a solid line; missing values remain gaps. Hover or use the keyboard for weekly units and event details, and expand the chart for a larger reading. The chart names its date window. In Commitment mode, reported observations outside that window are excluded.

Selecting a grid cell scopes the chart without changing saved quantities.

## What to do

1. Open **Forecast** and choose **Commitment** or a rolling 4, 13 or 26 weeks.
2. Choose **Doors**, select a **Product**, then narrow with **Find a door** and **Filters**. From an editable product cell, **Edit by door** opens that product's door view without saving an aggregate change.
3. Select a forecast cell for its P10, P50, P90, weekly-shape drivers, basis and events. The drawer distinguishes applied events from those only noted.
4. On one product's door view, choose **Week** for one week or **Row** for a percentage change across coming weeks. Enter a **Reason** to enable **Save**. Versions retain the engine's figure beside your edits. **Back to engine value** withdraws an edit.
5. Open **Product plans** to choose a saved product plan or start one. To fill that product's forecast from the engine, press **Edit forecast** on the product and then **Fill from engine**. It fills coming weeks from the run named beside it and states how many cells. Nothing saves until you save.

Only weeks that start after today can be edited. A cell on the product or category view is a sum of doors, and a cell on **All styles** is a sum of products: edits are made on one product's **Doors** view. A lift on a single door's event shows **Noted, not applied**.

## When it is empty or failed

- **Forecast unavailable** with **Retry**: the read failed. Retry it.
- **No forecast for this account**: the buyer has withheld consent. The button that asks the buyer for consent (it names the buyer) invites the buyer to the portal, and **Order from order history** opens the plan.
- **Forecast starts in 3 weeks**, over a meter of 5 of the 8 weeks reported. The meter fills as weeks arrive and the forecast starts at eight. Until then **Prepare order** offers **From comparable accounts** and **From order history**, and either opens the plan.
- **Recent reporting gap**: the recent account reports are not consecutive. The meter counts reported weeks in the latest four-week account window. **View reported sell-out** opens the reported history; missing weeks stay missing. This count does not promise that every product can be forecast.
- **No forecast for this account yet**: no run exists.
- **Not by door** under **Doors measured**: the account sends one total, so only **Styles** and **Categories** are offered and the forecast is by product and week.
