Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 5 (common margin): under independent-assortment search, hsi(vj)=πjsi(vj)−Lsi(vj)h^{si}(v_j)=\pi_j^{si}(v_j)-L^{si}(v_j)hsi(vj​)=πjsi​(vj​)−Lsi(vj​) is quasi-convex on [0,∞)[0,\infty)[0,∞)

Disproved
AssortSearch.SearchQC.theorem_5

by mikedeng1 · Oct 8, 2026 · Mathlib 0df444a (Lean v4.33.1)

assortment-planningconsumer-searchmultinomial-logitp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1quasi-convexity

Consider the independent-assortment search model of Cachon, Terwiesch and Xu. There are nnn variants with preferences vi>0v_i>0vi​>0 and a no-purchase option with preference v0>0v_0>0v0​>0; λ=exp⁡[−(Uˉ/μ+γ)]>0\lambda=\exp[-(\bar U/\mu+\gamma)]>0λ=exp[−(Uˉ/μ+γ)]>0 is the search parameter; every variant earns the same margin m∈Rm\in\mathbb Rm∈R; and the operational cost ccc is concave and increasing on [0,1][0,1][0,1] and twice continuously differentiable on (0,1)(0,1)(0,1). Variant iii's demand in an assortment SSS is

qisi(S)=vi∑k∈Svk+v0(1−e−λ(v0+∑k∈Svk)),q_i^{si}(S)=\frac{v_i}{\sum_{k\in S}v_k+v_0}\Bigl(1-e^{-\lambda(v_0+\sum_{k\in S}v_k)}\Bigr),qisi​(S)=∑k∈S​vk​+v0​vi​​(1−e−λ(v0​+∑k∈S​vk​)),

and its profit is πisi(S)=m qisi(S)−c(qisi(S))\pi_i^{si}(S)=m\,q_i^{si}(S)-c(q_i^{si}(S))πisi​(S)=mqisi​(S)−c(qisi​(S)).

Fix an assortment SSS and a variant j∉Sj\notin Sj∈/S, and give jjj a variable preference vj≥0v_j\ge0vj​≥0. Let πisi(vj)\pi_i^{si}(v_j)πisi​(vj​) be variant iii's profit in S∪{j}S\cup\{j\}S∪{j}, let Lsi(vj)=∑i∈Sπisi(S)−∑i∈Sπisi(vj)L^{si}(v_j)=\sum_{i\in S}\pi_i^{si}(S)-\sum_{i\in S}\pi_i^{si}(v_j)Lsi(vj​)=∑i∈S​πisi​(S)−∑i∈S​πisi​(vj​), and let

hsi(vj)=πjsi(vj)−Lsi(vj)h^{si}(v_j)=\pi_j^{si}(v_j)-L^{si}(v_j)hsi(vj​)=πjsi​(vj​)−Lsi(vj​)

be the change in total profit from adding jjj. Then hsih^{si}hsi is quasi-convex on [0,∞)[0,\infty)[0,∞): for all 0≤a≤b0\le a\le b0≤a≤b and t∈[a,b]t\in[a,b]t∈[a,b],

hsi(t)≤max⁡{hsi(a),hsi(b)}.h^{si}(t)\le\max\{h^{si}(a),h^{si}(b)\}.hsi(t)≤max{hsi(a),hsi(b)}.

Quasi-convexity means that adding a variant is most profitable at an extreme of its popularity, which is what places the optimal assortment among the popular assortments, the sets of the kkk most preferred variants.

Formalization Note The paper's model allows a margin mim_imi​ per variant. The theorem is stated with one margin mmm for every variant, including jjj, as the proof's displays (7)–(9) do; with unequal margins the statement is false (one variant i∈Si\in Si∈S with vi=0.76202736v_i=0.76202736vi​=0.76202736, v0=0.021273423v_0=0.021273423v0​=0.021273423, c(q)=0.38165308 q0.59833805c(q)=0.38165308\,q^{0.59833805}c(q)=0.38165308q0.59833805, mj=5.4080321m_j=5.4080321mj​=5.4080321, mi=6.1078441m_i=6.1078441mi​=6.1078441, λ=2.6494844\lambda=2.6494844λ=2.6494844 gives an hsih^{si}hsi that rises, falls and rises again). The paper assumes only "concave and increasing"; twice continuous differentiability on (0,1)(0,1)(0,1) is added because the proof uses c′c'c′ and c′′c''c′′. λ>0\lambda>0λ>0 is a free parameter (every real Uˉ\bar UUˉ gives one). Quasi-convexity is Mathlib's QuasiconvexOn ℝ (Set.Ici 0). At vj=0v_j=0vj​=0, hsi(0)=−c(0)h^{si}(0)=-c(0)hsi(0)=−c(0).

Preamble
import Mathlib
import Definitions.Def_AssortSearch_SearchQC_Model
Formal statement
namespace AssortSearch.SearchQC

/-- Theorem 5 (p. 14), common-margin case: under independent-assortment search, with one margin
`m` for every variant and a concave, increasing, twice continuously differentiable cost `c`,
`h^si(v_j) = π_j^si(v_j) - L^si(v_j)` is quasi-convex in `v_j` on `[0, ∞)`. `lam` is
`λ = exp[-(Ū/μ + γ)]`, taken as a free positive parameter. -/
theorem theorem_5 {n : ℕ} (v : Fin n → ℝ) (v0 lam m : ℝ) (c : ℝ → ℝ)
    (S : Finset (Fin n)) (j : Fin n)
    (hv : ∀ i, 0 < v i) (hv0 : 0 < v0) (hlam : 0 < lam) (hj : j ∉ S)
    (hconc : ConcaveOn ℝ (Set.Icc 0 1) c) (hmono : MonotoneOn c (Set.Icc 0 1))
    (hsmooth : ContDiffOn ℝ 2 c (Set.Ioo 0 1)) :
    QuasiconvexOn ℝ (Set.Ici 0) (hSI m c lam v v0 S j) := by sorry

end AssortSearch.SearchQC
Source
Cachon, Terwiesch & Xu, Retail Assortment Planning in the Presence of Consumer Search, working paper (Dec. 20, 2002), p. 14 (PDF 16), Theorem 5
Human review
  • Endorsed by Shuze Chen · Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 8, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me