Top-down vs bottom-up forecast reconciliation
Reconciles the category forecast (top-down) and the sum of the SKU forecasts (bottom-up): the scale, the bias of SKU forecasts and the reconciled SKUs at a given trust in the category level. Shows how much more optimistic the SKU forecasts are than the category.
At a glance
- What question does it answer?
- How much more optimistic the position forecasts are than the category and how to reconcile them with the chosen trust in the category.
- What data is needed?
- The category forecast.
- How to read the result?
- The bias shows the optimism of positions; pass the sum at the chosen trust to buying. Example on the default values: Sum of SKU forecasts, units: 1,150; Bias of the SKU forecasts to the category, %: +15%; Scale λ (category ÷ sum of SKUs): 0.8696; Sum after full reconciliation, units: 1,000. The figures are illustrative: enter your own to see your result.
Why and how to reason
The category forecast is steadier, position forecasts are more detailed, and the sum of positions is usually higher. In the example the category is 1,000 units, four SKUs give 1,150: a bias of +15.0%, a scale of 0.8696; at a trust of 50% the sum is 1,075 units.
When to use it
Apply when reviewing the forecast and before buying. The bias by period is counted by sopbias, forecast methods by forecast, here the reconciliation of levels.
How to read the result
The bias shows the optimism of positions; pass the sum at the chosen trust to buying.
Common mistakes
SKUs are edited by hand without reviewing the category, a trust of 100% is taken for one-off promotions.
What data is needed?
- The category forecast.
- The SKU forecasts.
- Trust in the category.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| Category forecast | pcs | 1,000 | The forecast for the whole category in the same units as the SKUs: top-down, usually steadier. The example value is illustrative. |
| Trust in the category level α | % | 50 | 100% means fitting the SKUs fully to the category, 0% keeping the SKUs as they are. Do not take α = 100% for a one-off category promotion. The example value is illustrative. |
How it is calculated
| Step | Formula |
|---|---|
| Bottom-up | B = the sum of the SKU forecasts |
| Scale and bias | λ = T ÷ B; bias = B ÷ T − 1 |
| Full reconciliation | SKU × λ |
| Compromise | SKU* = α × SKU × λ + (1 − α) × SKU, α = 50% |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Sum of SKU forecasts, units
- 1,150
- Bias of the SKU forecasts to the category, %
- +15 %
- Scale λ (category ÷ sum of SKUs)
- 0.8696
- Sum after full reconciliation, units
- 1,000
- Sum after reconciliation with trust α, units
- 1,075
- Change of the SKU sum at α, units
- −75
- Bottom-up
- Full reconciliation
- With trust α
Data as a table
| Measure | Bottom-up | Full reconciliation | With trust α |
|---|---|---|---|
| SKU 1 | 400 | 348 | 374 |
| SKU 2 | 350 | 304 | 327 |
| SKU 3 | 250 | 217 | 234 |
| SKU 4 | 150 | 130 | 140 |
SKU reconciliation
| SKU | Bottom-up | Full reconciliation | With trust α | Change |
|---|---|---|---|---|
| SKU 1 | 400 | 347.8 | 373.9 | −26.1 |
| SKU 2 | 350 | 304.3 | 327.2 | −22.8 |
| SKU 3 | 250 | 217.4 | 233.7 | −16.3 |
| SKU 4 | 150 | 130.4 | 140.2 | −9.8 |
The SKU forecasts add up above the category by 15 %: a typical optimism of position forecasts. Find the cause, do not only scale.
Category forecasts are steadier when sales per position are small; the reconciled sum at α between 0 and 100% lies between bottom-up and top-down.
What to do with the result: edit SKUs together with the review of the category, not by hand one by one; pass the reconciled values to the size split and to buying.
A typical mistake: editing SKUs by hand without reviewing the category and taking α = 100% for one-off promotions.
Nuances and limits
- Editing SKUs by hand without reviewing the category.
- Taking α = 100% for one-off promotions.
- Mixing units and periods.