Mystery Box Manager
produce allocation engine
Every week, surplus produce needs distributing into mystery boxes.
By hand, it's slow and uneven. Some boxes get shortchanged.
An ILP solver balances value, variety, and fairness at once.
Six penalty dimensions. 800 tuning trials. Every box is fair.
Tweak the priorities. The algorithm adapts instantly.
scroll to explore ▿
ILP

Pipeline

  • Volatile prices, seasonal variety and unpredictable surplus create a fresh problem each week
  • ILP solver finds provably optimal allocations across multiple box tiers
  • Multi-objective optimisation for value, diversity, fairness and more
  • Validated against years of real historical packing data
62 45 38 71 80 76 Manual 71 α β γ δ ε ζ

Hyper-dimensional scoring

  • Multiple penalty dimensions score every box: value, diversity, fungibility and more
  • ILP penalty scores embed power curves with piecewise-linear tangent lines
  • HHI concentration measures effective diversity per dimension
  • Two-layer model caps flooding at individual and fungible group levels
71 score

Tuning to proof

  • Bayesian optimisation tunes scoring parameters against years of packing history
  • Cross-validated with repeated-run stability grading per parameter
  • Glass-box EBM diagnoses scoring gaps and distils new penalty terms
  • HITL preference ratings tune penalties via Spearman correlation