Lemma 2, p. 13 — one extra past purchase costs at most the highest remaining price:
ProvedAssortRotation.Concealment.lemma_2Products are offered sequentially; a consumer who has already bought items buys the current product iff , where is nonnegative and decreasing. Let be the revenue from the last products when the consumer enters them having bought items. Let be the highest price among those products, with . If all prices are nonnegative, then for every
A consumer with one more past purchase is less willing to buy. She nonetheless buys at most one product fewer, so she costs the retailer at most one remaining price. This bound drives the interchange argument (Lemma 4) behind Theorem 1.
Formalization Note. The last products are an arbitrary list of (price, shock) pairs, and the statement holds for every realization. The paper's "with probability 1" is thus strengthened to "always". The paper restricts to ; the Lean statement holds for every . The purchase rule uses (as in (5)–(6)) where (7) prints . The printed " if " is read as , with .
import Mathlib import Definitions.Def_AssortRotation_Concealment_Model
namespace AssortRotation.Concealment
/-- Lemma 2, p. 13. For any list `L` of products (the last `n` products of the sequence) with
nonnegative prices and any number `k` of past purchases,
`R_{k+1}(n) ≤ R_k(n) ≤ ρ*_n + R_{k+1}(n)`, where `R_k(n) = seqRev v L k` and
`ρ*_n = rhoStar L` is the highest remaining price (`0` for `n = 0`). -/
theorem lemma_2 (v : ℕ → ℝ) (hv : MarginalUtility v) (L : List (ℝ × ℝ))
(hp : ∀ a ∈ L, 0 ≤ a.1) (k : ℕ) :
seqRev v L (k + 1) ≤ seqRev v L k ∧ seqRev v L k ≤ rhoStar L + seqRev v L (k + 1) := by sorry
end AssortRotation.Concealment
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.