The engine

Sometimes you know something the model does not.

A launch with no history. A line you are discontinuing. A number the board already committed to. Set it by hand, and keep it next to the statistical forecast rather than on top of it — so you can see exactly where the two disagree.

For the SKUs with no history to fit

The engine needs about a year of sales before it can fit a seasonal pattern. A launch has none. A manual plan gives that SKU a real number to buy against in the meantime.

Next to the forecast, not instead of it

Both numbers stay visible. When the statistical forecast catches up and starts disagreeing with your assumption, you find out — which is the whole point of writing the assumption down.

The board number is a plan too

A commitment made in a board meeting drives real purchasing whether or not the model agrees with it. Keeping it as an explicit plan makes the gap between ambition and demand a number instead of an argument.

Overnight

What the agent does with your own numbers

Every night

The agent plans against your manual number where you set one, and the overnight check tells you when actual sales have drifted away from it.

What it can’t do

It cannot set, change, or override a manual plan. Where you have said what the number is, that is the number.

The launch with nothing to forecast from

A new flavour ships in six weeks with zero history. The model has nothing to fit and says so. The team sets a manual plan from comparable launches and buys against it — and eight weeks in, with real sales on the board, the statistical forecast starts disagreeing. That disagreement is the signal to re-buy, and it only exists because the assumption was written down.

Plan the SKUs with no history

Bring your next launch and see it planned beside the forecast, not against it.