Stock movements

How to make forecasts made at different levels ADD UP with each other. Each area forecasts its part, management forecasts the total, and the two never match — here the difference is distributed with an explicit criterion instead of by decree.

The sales/revenue forecast in a HIERARCHY (SKU → customer → region → total) reconciled so that the levels ADD UP exactly. A forecast that does not add up is the no. 1 reason for the owner's distrust: if the sum of the SKUs does not match the total, which is the true number? Here you see the RAW value (forecast in isolation) next to the RECONCILED one, with the adjustment in plain sight, by two methods: BOTTOM-UP (the base rules — the sum of the leaves defines the parents) and PROPORTIONAL TOP-DOWN (the top rules — the total is distributed by the leaves' weights). In both, Σ children = parent at every level, with the rounding residue handled (the last child takes the remainder) — not a single cent is lost.

Bottom-up: the sum of the parts becomes the total (it trusts those on the front line). Proportional top-down: the total sent from above is distributed among the parts in the proportion they already had (it trusts the aggregate view). The choice is political, not technical — and the screen shows the effect of each before applying.

Adjustment is the difference between the reconciled value and the raw value of each node — signed, highlighted when it pulled downward. It is the column that answers "who gave up how much": in bottom-up the leaves have zero adjustment (they rule) and the parents absorb the difference; in proportional top-down the total has zero adjustment and the leaves share the allocation by weight, with the rounding residue going to the last child — not a single cent is lost along the way.

The "Restore example" button brings back the demonstration hierarchy with the original values — useful for experimenting with both methods and seeing how the adjustment behaves without committing anything. Since the leaves are editable, you can simulate: zero out a leaf, reconcile in bottom-up and watch the parents shrink; then switch to proportional top-down and watch the same total be redistributed by the weights of the remaining leaves.

It is the screen's guarantee: at any point in the tree, the sum of the children equals the parent. The RAW column keeps what each one originally forecast, so you can see how much each area was adjusted — the difficult conversation comes with evidence.

Does my forecast ADD UP — does the sum of the SKUs match the total? Choose the method (bottom-up or proportional top-down) and click Reconcile: the panel declares "Adds up" with the reconciled total, and the table shows, node by node, the raw value (forecast in isolation), the reconciled one and the adjustment applied. On the leaves the raw value is editable — change it and reconcile again to see the effect rise through the hierarchy.