Key concepts in 5 minutes
This page covers the minimum vocabulary you need to understand KPIs, the collaborative cycle, and the inventory module. For complete definitions and formulas, see the Bilingual Glossary.
Forecast error
In AInventory, error is always defined as:
error = D − F
where D is the observed actual demand and F is the forecast for that period.
The sign convention is intentional:
| Error sign | Interpretation | Main risk |
|---|---|---|
Positive (D > F) | Actual demand exceeded the forecast | Stockout |
Negative (D < F) | Forecast exceeded actual demand | Overstock / holding cost |
| Zero | Perfect forecast | — |
A persistently positive error on a critical SKU means the forecasting process is systematically underestimating demand — the highest-risk operational scenario. A negative error means tied-up capital. Both carry a cost, but with different impacts on fill rate.
Forecast vs actual demand
- Forecast — the demand projection for a future period, produced by one of the four phases of the collaborative cycle.
- Actual demand — the quantity actually sold or consumed once the period has closed. This is the value used to compute the error and measure KPIs.
AInventory can only compute KPIs when the period has already elapsed and the version is closed (see below).
The phases of the collaborative cycle
The forecasting process moves through four sequential phases. Each one produces its own forecast for the same planning horizon:
- Baseline (AI) — generated automatically by the statistical engine.
- Sales — adjusted by the commercial team.
- Marketing — enriched with marketing plans and promotions.
- Consensus — agreed version across Supply Chain, Sales, and Marketing; used as the operational input.
Open version vs closed version
Each planning cycle produces a version of the forecast. The version's status determines what can be done with it:
| Status | Description | Feeds KPIs and FVA? |
|---|---|---|
| Open | The version is under active editing; values can be changed. | No |
| Closed | The cycle has ended; values are locked. | Yes, when the period has also elapsed |
KPIs and FVA (Forecast Value Added) are only calculated on closed versions whose periods have already passed. A future period or an open version has no measurable error.
KPIs: three levels of aggregation
AInventory reports KPIs at three levels that coexist in the same view:
| Level | What it measures |
|---|---|
| Per SKU | Forecast performance for one product in a specific version/period. |
| Per version | Average of all SKU-level KPIs in that version (each SKU weighted equally). |
| Cumulative | Historical average across all closed versions of the tenant. |
The equal weight per SKU (not volume-weighted) means a slow-moving product has the same influence on the version-level indicator as a top seller. This incentivizes forecast quality across the entire portfolio.
Fill rate and factor k (inventory)
These two concepts belong to the Inventory Optimization module:
- Fill rate — the probability of satisfying a SKU's demand without a stockout during the lead time. Expressed as a percentage (e.g., 95 %).
- Factor k — the number of demand standard deviations added to the cycle stock to achieve the target fill rate. AInventory determines it via Monte Carlo simulation, minimizing total expected cost (holding + shortage).
With these concepts in mind, you are ready to navigate the interface. Continue with Sign in and access or check the Bilingual Glossary for additional terms.