Lemma D.1 on renewal windows: a shortened window contracts the mirror cost by and does not raise holding and backlog cost
ProvedOWMR.dfr_window_contractionapproximation-algorithmsinventoryinventory-theoryoperations-researchrenewal-theorystochastic-control
Fix retailer of a network under the standing assumptions, with renewal measure of its demand. Let and satisfy , and suppose is concave on . Then:
- (mirror contraction) for all reals and , ;
- (holding and backlog) for every there is a target with ,
where is the mirror cost (14) and (ssAvgCost) is the long-run average holding and backlog cost of the policy. The long-run depletion of the width- cycle retains the lowest fraction of the probability mass of the width- depletion, so these are the inequalities (D.2) and the outer inequality of (D.3).
Preamble
import Definitions.Def_owmr_renewal open MeasureTheory open scoped ENNReal NNReal
Formal statement
namespace OWMR
theorem dfr_window_contraction {n : ℕ} (N : Network n) (hN : N.Standing) (i : Fin n)
(wh w t : ℝ) (hw : 0 < w) (hwwh : w ≤ wh) (ht0 : 0 < t) (ht1 : t ≤ 1)
(hM : renewalM (N.marginal i) w = ENNReal.ofReal t * renewalM (N.marginal i) wh)
(hconc : ConcaveOn ℝ (Set.Icc 0 wh)
(fun x => (renewalMeasure (N.marginal i) (Set.Icc 0 x)).toReal)) :
(∀ S Sh : ℝ, mirrorCost (N.marginal i) N.h N.p (S - w) S
≤ ENNReal.ofReal t * mirrorCost (N.marginal i) N.h N.p (Sh - wh) Sh) ∧
∀ Sh : ℝ, ∃ S : ℝ, ssAvgCost (N.marginal i) (N.F i) 0 (S - w) S
≤ ssAvgCost (N.marginal i) (N.F i) 0 (Sh - wh) Sh := by sorry
end OWMRSource
M. Hu, "A 5/2 Guarantee for Echelon Stock (s,S) Policies in One-Warehouse Multi-Retailer Systems", working paper (manuscript of November 2026), Appendix D, Lemma D.1 (Convex-Quantile Contraction), (D.2) and (D.3), p. E.25, proved in Appendix D.1, pp. E.27-E.28; applied in Step 2 of the proof of Theorem D.1, p. E.26, including its footnote on the three cost terms.