Optimizing Li Auto Store Closures with Scoring and 0‑1 Integer Programming

Background and Problem

In January 2026, Li Auto intends to close roughly 100 retail stores in the first half of the year, about 18% of its directly‑operated network. After expanding from 206 stores in 2021 to 548 in 2025 (a 166% increase), deliveries fell to 406,000 units in 2025, a 19% YoY drop, and the company posted a net loss of ¥624 million.

Store efficiency also deteriorated: average annual deliveries per store fell from 997 in 2024 to 741 in 2025, roughly two cars per day. Operating costs are high, with prime‑city stores paying ¥2‑3.6 million in rent and total annual costs around ¥5 million per store.

Store Scoring Model

To identify “low‑efficiency” stores, a weighted scoring system is constructed from three dimensions.

Sales Efficiency (40%)

Measured by monthly sales and conversion rate:

S = 0.6 × (Monthly Sales / Avg Monthly Sales) + 0.4 × (Conversion Rate / Avg Conversion Rate)

Example: a store with 30 cars/month and 6% conversion (industry avg 15%) yields S = 0.45.

Cost Efficiency (35%)

Based on annual operating cost versus gross profit:

C = 1 - (Annual Operating Cost / Annual Gross Profit)

For a flagship store: cost ¥5 million, annual sales 360 cars, profit per car ¥20,000, giving annual gross profit ¥7.2 million and C = 0.31.

Strategic Value (25%)

Qualitative factors include geographic uniqueness, brand exposure, and market potential (e.g., local new‑energy vehicle penetration).

The overall score is:

Total Score = 0.4×S + 0.35×C + 0.25×V

Cost‑Benefit Analysis

Assuming the 100 lowest‑scoring stores are closed, the following assumptions are used:

Annual operating cost per closed store: ¥4 million

Annual sales per closed store: 300 cars

Profit per car: ¥20,000

Annual gross profit per closed store: ¥6 million

Direct benefits: saved operating cost ¥4 million × 100 = ¥400 million and avoided gross‑profit loss ¥6 million × 100 = ¥600 million, netting roughly ¥3 billion.

Hidden cost: potential 5% loss of overall sales (≈2.03 万 cars) translates to ¥40.6 million × 5% ≈ ¥2.03 万 cars lost, or ¥40.6 million in gross profit, reducing net benefit to about ¥3 billion.

Optimization Model

The closure decision is formulated as a 0‑1 integer programming problem.

Decision variable : x_i = 1 if store i is closed, 0 otherwise.

Objective : Maximize net benefit = Σ (Benefit_i × x_i) – Σ (HiddenCost_i × x_i).

Constraints :

Exactly 100 stores must be closed: Σ x_i = 100.

Each prefecture‑level city retains at least one store.

Service radius (≈50 km) must still cover the original market.

Stores with total score < 0.4 are prioritized for closure.

Prioritized closures :

Low‑efficiency flagship stores in prime city districts (rent > ¥3 million/yr, sales < 5 cars/month).

Redundant stores within 10 km of another store in the same city.

Low‑sales stores in third‑ and fourth‑tier cities (new‑energy penetration < 15%, annual sales < 200 cars).

AutoPark‑type stores that can be replaced by nearby supermarkets.

Must‑keep stores :

Only store in a region.

Stores in cities with new‑energy penetration > 40%.

High‑efficiency stores with annual sales > 1,000 cars.

Applying this strategy yields estimated cost savings of ¥2.93 billion, sales loss of about 25,000 cars (≈6%), and a net benefit of ¥2.43 billion (or ¥3 billion under conservative profit assumptions). This could offset 27‑39% of the ¥6.24 billion loss reported for 2025.

Issues Beyond the Model

Product‑Line Dilemma

Li Auto’s six current models are all mid‑large or large SUVs, causing severe internal competition. For a consumer with a ¥350,000 budget, the choice among L7, L8, and i8 often defaults to the cheapest (L7), cannibalizing sales of higher‑priced variants. After L7’s launch, L8’s monthly sales fell from 15,000 to 8,000 units; L6’s launch further eroded L7’s market.

Competitive Landscape

¥450‑500k segment: L9 vs. WM M9 (M9 leads with Huawei’s autonomous driving).

¥350‑400k segment: L8 vs. WM M8 (M8 quickly reduced L8 sales).

¥300‑350k segment: L7 vs. WM M7 (M7 offers more aggressive pricing and features).

¥250‑300k segment: L6 vs. Xiaomi YU7, Lynk & Co 900 (price‑sensitive competition).

Li Auto lacks a dominant advantage in any price band, while rivals benefit from strong tech backing or brand appeal.

Conclusion

From an operations research perspective, a scoring‑based and integer‑programming approach can quantify store‑closure decisions and potentially save ¥2‑3 billion, easing cash‑flow pressure. However, the fundamental challenge lies in Li Auto’s over‑concentrated SUV product line and fierce competition, which must be addressed to ensure long‑term viability of the remaining 448 stores.