Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 3, pp. 1385–1386 — one markdown or markup: strictly increasing thresholds x_n < t¹_n < t_n < t²_n < z_n with J(n,t) = V(n,t)

Disproved
PriceSwitch.UpOrDown.theorem3_markup_or_markdown_thresholds

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

dynamic-pricingmarkdownmarkupoptimal-stoppingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1threshold-policy

Sales start at price ppp with Poisson demand of intensity λ\lambdaλ, and once during the horizon the price may be changed either down to p1p_1p1​ or up to p2p_2p2​, where

p1<p<p2,λ1>λ>λ2,r1=p1λ1>r=pλ>r2=p2λ2,p_1 < p < p_2, \qquad \lambda_1 > \lambda > \lambda_2, \qquad r_1 = p_1\lambda_1 > r = p\lambda > r_2 = p_2\lambda_2,p1​<p<p2​,λ1​>λ>λ2​,r1​=p1​λ1​>r=pλ>r2​=p2​λ2​,

all positive. Let Ji(n,t;0)J^i(n,t;0)Ji(n,t;0) be the revenue of switching at once to pip_ipi​, J(n,t;0)=max⁡{J1(n,t;0),J2(n,t;0)}J(n,t;0) = \max\{J^1(n,t;0), J^2(n,t;0)\}J(n,t;0)=max{J1(n,t;0),J2(n,t;0)}, J(n,t;t)J(n,t;t)J(n,t;t) the revenue of keeping ppp, and J(n,t)J(n,t)J(n,t) the optimal expected revenue over all non-anticipating switching times. Let Gi(n,t)=r−ri−pi(λ−λi)P(Ni(t)≥n)G^i(n,t) = r - r_i - p_i(\lambda - \lambda_i)P(N_i(t) \ge n)Gi(n,t)=r−ri​−pi​(λ−λi​)P(Ni​(t)≥n), Δi(n,t)=J(n,t;t)−Ji(n,t;0)\Delta^i(n,t) = J(n,t;t) - J^i(n,t;0)Δi(n,t)=J(n,t;t)−Ji(n,t;0), Δ(n,t)=J1(n,t;0)−J2(n,t;0)\Delta(n,t) = J^1(n,t;0) - J^2(n,t;0)Δ(n,t)=J1(n,t;0)−J2(n,t;0), tni=inf⁡{t>0:Δi(n,t)=0}t^i_n = \inf\{t > 0 : \Delta^i(n,t) = 0\}tni​=inf{t>0:Δi(n,t)=0} and tn=inf⁡{t>0:Δ(n,t)=0}t_n = \inf\{t > 0 : \Delta(n,t) = 0\}tn​=inf{t>0:Δ(n,t)=0}.

Assumption (as on p. 1385): tn1<tn<tn2t^1_n < t_n < t^2_ntn1​<tn​<tn2​ for every n≥1n \ge 1n≥1.

Theorem. There exist two strictly increasing sequences (xn)n≥1(x_n)_{n\ge1}(xn​)n≥1​ and (zn)n≥1(z_n)_{n \ge 1}(zn​)n≥1​ of time thresholds with

xn<tn1<tn<tn2<zn,x_n < t^1_n < t_n < t^2_n < z_n,xn​<tn1​<tn​<tn2​<zn​,

such that J(n,t)=V(n,t)J(n,t) = V(n,t)J(n,t)=V(n,t) for all nnn and t≥0t \ge 0t≥0, where V(0,t)=0V(0,t) = 0V(0,t)=0 and, for n≥1n \ge 1n≥1,

V(n,t)={J(n,t;0)+F(n,t)if xn≤t≤zn,J(n,t;0)otherwise,F(n,t)={H1(n,t)if xn≤t≤tn,H2(n,t)if tn≤t≤zn,0otherwise,V(n,t) = \begin{cases} J(n,t;0) + F(n,t) & \text{if } x_n \le t \le z_n, \\ J(n,t;0) & \text{otherwise,} \end{cases} \qquad F(n,t) = \begin{cases} H^1(n,t) & \text{if } x_n \le t \le t_n, \\ H^2(n,t) & \text{if } t_n \le t \le z_n, \\ 0 & \text{otherwise,} \end{cases}V(n,t)={J(n,t;0)+F(n,t)J(n,t;0)​if xn​≤t≤zn​,otherwise,​F(n,t)=⎩⎨⎧​H1(n,t)H2(n,t)0​if xn​≤t≤tn​,if tn​≤t≤zn​,otherwise,​

with F(0,t)=0F(0,t) = 0F(0,t)=0. Here, with Li(n,t)=Gi(n,t)+λ[V(n−1,t)−Ji(n−1,t;0)]L^i(n,t) = G^i(n,t) + \lambda[V(n-1,t) - J^i(n-1,t;0)]Li(n,t)=Gi(n,t)+λ[V(n−1,t)−Ji(n−1,t;0)]:

  1. H1(n,⋅)H^1(n,\cdot)H1(n,⋅) solves ∂H/∂t=−λH+L1(n,t)\partial H/\partial t = -\lambda H + L^1(n,t)∂H/∂t=−λH+L1(n,t) with H(n,xn)=0H(n,x_n) = 0H(n,xn​)=0;
  2. H2(n,⋅)H^2(n,\cdot)H2(n,⋅) solves ∂H/∂t=−λH+L2(n,t)\partial H/\partial t = -\lambda H + L^2(n,t)∂H/∂t=−λH+L2(n,t) with H(n,tn)=H1(n,tn)H(n,t_n) = H^1(n,t_n)H(n,tn​)=H1(n,tn​);
  3. xn=inf⁡{t≥0:L1(n,t)=0}x_n = \inf\{t \ge 0 : L^1(n,t) = 0\}xn​=inf{t≥0:L1(n,t)=0} and zn=inf⁡{t≥tn:H2(n,t)≤0}z_n = \inf\{t \ge t_n : H^2(n,t) \le 0\}zn​=inf{t≥tn​:H2(n,t)≤0}, both sets being nonempty.

In words: with nnn items left, it is optimal to mark down to p1p_1p1​ immediately when the time-to-go is below xnx_nxn​, to mark up to p2p_2p2​ immediately when it is above znz_nzn​, and in between to keep ppp, the stopping value exceeding the switch-at-once revenue by F(n,t)F(n,t)F(n,t). This is the main result of §5.

Formalization Note The sequences, FFF, H1H^1H1, H2H^2H2 and VVV are existentially quantified; VVV is tied to FFF and FFF to H1,H2H^1, H^2H1,H2 by explicit equations, each Hi(n,⋅)H^i(n,\cdot)Hi(n,⋅) is pinned by its linear ODE (imposed at every t>0t > 0t>0) and its boundary value, and xn,znx_n, z_nxn​,zn​ by their defining infima, so the existential determines the paper's objects; VVV is a separate witness only because Li(n,⋅)L^i(n,\cdot)Li(n,⋅) reads V(n−1,⋅)V(n-1,\cdot)V(n−1,⋅). At n=1n = 1n=1, V(0,t)−Ji(0,t;0)=0V(0,t) - J^i(0,t;0) = 0V(0,t)−Ji(0,t;0)=0, as the page's "J(n,t;0)≐0J(n,t;0) \doteq 0J(n,t;0)≐0 for n≤0n \le 0n≤0" gives. The page's closed intervals overlap at tnt_ntn​, where the two branches of FFF agree by the boundary condition of H2H^2H2; the formalization keeps the page's inequalities with the first branch taking precedence. The page's assumption tn1<tn<tn2t^1_n < t_n < t^2_ntn1​<tn​<tn2​ is a hypothesis, as on the page; the three infima are Lean sInfs, whose sets are nonempty by Lemma 3. In §5 the low price is p1p_1p1​ and the high price p2p_2p2​, with initial price ppp (in §4, p1p_1p1​ was the initial price). The demand at ppp is the referenced exponential-interarrival counting process (IsExpInterarrivals P lam T); IsProbabilityMeasure P is redundant given it. The sum of independent Poisson counts in Ji(n,t;s)J^i(n,t;s)Ji(n,t;s) is written through its Poisson law.

Preamble
import Mathlib
import Definitions.Def_PriceSwitch_UpOrDown_ThreePrice
Formal statement
namespace PriceSwitch.UpOrDown

open MeasureTheory ProbabilityTheory QueueingFundamentals.Foundations

/-- Feng–Gallego (1995), Theorem 3, pp. 1385–1386: with one markdown (to `p₁`) or one markup (to `p₂`)
allowed from the initial price `p`, and under the page's assumption `t¹_n < t_n < t²_n`, there are strictly
increasing thresholds `x_n < t¹_n < t_n < t²_n < z_n` with `J(n, t) = V(n, t)`. -/
theorem theorem3_markup_or_markdown_thresholds
    {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P] (T : ℕ → Ω → ℝ)
    (p₁ lam₁ p lam p₂ lam₂ : ℝ)
    (h3 : IsThreePrice p₁ lam₁ p lam p₂ lam₂)
    (hT : IsExpInterarrivals P lam T)
    (hassume : ∀ n : ℕ, 1 ≤ n →
      tThr1 p₁ lam₁ p lam p₂ lam₂ n < tThr p₁ lam₁ p lam p₂ lam₂ n ∧
      tThr p₁ lam₁ p lam p₂ lam₂ n < tThr2 p₁ lam₁ p lam p₂ lam₂ n) :
    ∃ (x z : ℕ → ℝ) (F H₁ H₂ V : ℕ → ℝ → ℝ),
      let L₁ : ℕ → ℝ → ℝ := fun n t =>
        PriceSwitch.Markdown.G p lam p₁ lam₁ n t + lam * (V (n - 1) t - PriceSwitch.Markdown.switchRevenue p lam p₁ lam₁ (n - 1) t 0)
      let L₂ : ℕ → ℝ → ℝ := fun n t =>
        PriceSwitch.Markdown.G p lam p₂ lam₂ n t + lam * (V (n - 1) t - PriceSwitch.Markdown.switchRevenue p lam p₂ lam₂ (n - 1) t 0)
      StrictMonoOn x (Set.Ici 1) ∧ StrictMonoOn z (Set.Ici 1) ∧
      (∀ n, 1 ≤ n →
        x n < tThr1 p₁ lam₁ p lam p₂ lam₂ n ∧
        tThr1 p₁ lam₁ p lam p₂ lam₂ n < tThr p₁ lam₁ p lam p₂ lam₂ n ∧
        tThr p₁ lam₁ p lam p₂ lam₂ n < tThr2 p₁ lam₁ p lam p₂ lam₂ n ∧
        tThr2 p₁ lam₁ p lam p₂ lam₂ n < z n) ∧
      (∀ t, V 0 t = 0) ∧
      (∀ n, 1 ≤ n → ∀ t, V n t =
        if x n ≤ t ∧ t ≤ z n then switchNowMax p₁ lam₁ p lam p₂ lam₂ n t + F n t
        else switchNowMax p₁ lam₁ p lam p₂ lam₂ n t) ∧
      (∀ t, F 0 t = 0) ∧
      (∀ n, 1 ≤ n →
        (∀ t, 0 < t → HasDerivAt (H₁ n) (-lam * H₁ n t + L₁ n t) t) ∧
        H₁ n (x n) = 0 ∧
        (∀ t, 0 < t → HasDerivAt (H₂ n) (-lam * H₂ n t + L₂ n t) t) ∧
        H₂ n (tThr p₁ lam₁ p lam p₂ lam₂ n) = H₁ n (tThr p₁ lam₁ p lam p₂ lam₂ n) ∧
        (∃ t, 0 ≤ t ∧ L₁ n t = 0) ∧
        x n = sInf {t : ℝ | 0 ≤ t ∧ L₁ n t = 0} ∧
        (∃ t, tThr p₁ lam₁ p lam p₂ lam₂ n ≤ t ∧ H₂ n t ≤ 0) ∧
        z n = sInf {t : ℝ | tThr p₁ lam₁ p lam p₂ lam₂ n ≤ t ∧ H₂ n t ≤ 0} ∧
        (∀ t, F n t =
          if x n ≤ t ∧ t ≤ tThr p₁ lam₁ p lam p₂ lam₂ n then H₁ n t
          else if tThr p₁ lam₁ p lam p₂ lam₂ n ≤ t ∧ t ≤ z n then H₂ n t
          else 0)) ∧
      (∀ n t, 0 ≤ t → optRevenue3 P T p₁ lam₁ p lam p₂ lam₂ n t = V n t) := by sorry

end PriceSwitch.UpOrDown
Source
Feng & Gallego (1995), Management Science 41(8), Theorem 3, pp. 1385–1386 (assumption t¹_n < t_n < t²_n, p. 1385; proof: Appendix, pp. 1388–1390)
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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