Validation and backing (VN2)
The backing of the paradigm: the VN2 Inventory Planning Challenge (2025-2026), where the winning approach cut total cost by 13% versus the reference method.
VN2 is the VN2 Inventory Planning Challenge (2025-2026), an external international competition cited as validation of the paradigm — it is not a policy or a product module. See ESTRUCTURA.md §5.3.
What the VN2 Inventory Planning Challenge is
The VN2 Inventory Planning Challenge (2025-2026) is an international inventory planning competition, organized independently of Directrix and AInventory. Participating teams receive real historical data and must produce replenishment plans; their results are evaluated against a reference method on the same dataset.
The VN2 is relevant to AInventory not as a product feature, but as external evidence that the predict-then-optimize + simulation paradigm outperforms traditional inventory formulas under real-world conditions.
The key result
The winning approach in the competition reduced total cost by approximately 13% versus the reference method. That 13% does not come from fine-tuning parameters or a data advantage: it comes from a paradigm shift.
The most significant finding is not the exact figure, but its cause: no winning team used the classic safety stock formula (z · σ_d · √L). The top-performing teams converged on variants of the predict-then-optimize approach — generate scenarios of future demand, evaluate policies over those scenarios, and select by expected total cost minimization — which is exactly the paradigm AInventory implements.
Why VN2 is relevant evidence
The standard objection to inventory optimization models is that they work well in theory but are fragile in practice: assumptions fail, data is noisy, planners do not trust black boxes. The VN2 addresses that objection directly:
- Data is real, not synthetic.
- Participants are independent teams, not the software vendor.
- Evaluation is objective: total cost on the same dataset.
In that context, a 13% reduction versus a reference method — which typically includes classic safety stock formulas and fixed coverage — is evidence that the paradigm generates measurable value.
What VN2 does not imply
The VN2 is an academic/professional competition with controlled conditions. Some limits of that evidence:
- Results depend on the specific competition dataset; generalization to any industry or product category requires caution.
- The 13% reduction is the result of the winning approach, not a typical expected result for any implementation.
- AInventory implements the VN2 paradigm but with its own design decisions (four-level cascade, five policies, specific cost model). The VN2 figure supports the paradigm, not a particular implementation.
VN2 is an external reference validating the predict-then-optimize paradigm, not a certification of the AInventory product. Its value is showing that the technical direction is correct, not guaranteeing a 13% reduction in any specific case.