Appendices
The trip and our sources
A delegation of some twenty French industrial leaders spent a week in the region, from 26 to 30 October 2026. It visited Honor, H&T, BYD, AION, XPENG, Shein's innovation centre, Vipshop and Huawei, as well as SCUT university in Guangzhou (see chapter 1).
Our sources are visit notes, interviews with the delegation's manufacturers and the public data cited in the notes.
Figures given by companies during the visits are reported as announced. Appendix C sets out the status of each one.
- Figures given by the companies visited are attributed to their source (“according to Shein”, “according to Huawei”). We noted them without auditing them.
- The visits were tightly controlled: a site tour, then a presentation in Chinese with live interpretation. We had few chances to get into the detail of the “how”. Our findings describe what we were shown and what we saw on the lines.
- The manufacturers' reactions are reported on behalf of the delegation, without individual attribution.
The delegation members
The delegation numbered some twenty people. The table lists fourteen industrial leaders and the three OSS Ventures members who accompanied them. It shows their role, their sector and the size of their company, based on the latest annual reports and press releases. The industrial companies are not named. Participants' reactions are reported on behalf of the delegation, never individually.
| Participant | Role | Sector | Revenue | Headcount | Sites |
|---|---|---|---|---|---|
| Aude De Menthières | Technical Director | Automotive glazing (business of a materials group) | €46.5bn (group, 2025) | About 161,000 (group) | Nearly 800 industrial sites (group), 80 countries |
| Anne-Catherine Huillery | Director of Digital Transformation & AI Operations | Pharmaceuticals and dermo-cosmetics | €3.2bn (2025) | About 10,200 | 6 production sites in France, nearly 90% of production in France |
| Wesley Lidy | Director 4.0 | Cosmetics | About €1.5bn | More than 8,000 | 3 industrial sites in France, presence in 150 countries |
| Julien Vialle | Director 4.0 | Semiconductors (materials) | About €0.6bn (financial year ended March 2026) | About 2,000 | 2 production sites (France, Singapore) |
| Julien Petibon | IT Manager | ||||
| Antoine Bois | COO | Railway infrastructure | €1.34bn (2025) | About 5,500 | More than 60 production sites in 21 countries |
| Serge Hauert | Manufacturing Organization Director | Luxury jewellery and watchmaking | Not published (group's jewellery houses: €16.5bn) | More than 9,000 | 8 jewellery sites and 5 watchmaking sites |
| Jérémie Aubert | Operational Excellence Director | ||||
| Emmanuelle Boisseau | Operational Excellence Director | Home automation and motorisation | €1.54bn (2025) | About 6,500 | 8 production sites, presence in 58 countries |
| Anthony Salapete | VP Operations | Threads and components for apparel and footwear | $1.46bn (2025) | More than 16,000 | Operations in some fifty countries |
| Philippe Seille | Studio Director | Sporting goods (entity of a design and retail group) | €16.8bn (group, 2025) | 103,000 (group) | 1,902 stores in 82 countries |
| François Delaunoy | Project Lead, Biomaterials | ||||
| Hubert Cariou | CIO | Packaging for beauty and personal care | About $1.2bn (2021, unlisted company) | More than 10,000 | 34 industrial sites |
| Jean-Philippe Ollier | VP Engineering | Tyres | €26.0bn (2025) | About 132,000 | 121 production sites in 26 countries |
| Renan Devillieres | Founder and CEO | OSS Ventures, industrial venture studio accompanying the delegation | Not applicable | 20 | Portfolio of 25 software startups deployed on more than 3,800 industrial sites |
| Jean-Philippe Lorinquer | Partner and Chief Revenue Officer | ||||
| Ashok Azhagarasan | Head of Partnerships |
Sources: 2025-2026 annual reports and press releases of the groups. References available on request.
Comparison between China and the West
Three public benchmarks help place the gap. They cover different scopes and should be read as orders of magnitude.
| Indicator | China | Europe | United States | Source |
|---|---|---|---|---|
| Monthly wage of a worker | About €900 (2025 average) | Gross minimum wage from €620 (Bulgaria) to €2,704 (Luxembourg); €1,823 in France, €2,343 in Germany (January 2026) | Federal minimum wage of $7.25 an hour, about $1,260 a month, and up to $17.13 an hour in Washington State, about $2,970 (2026). Average hourly earnings of manufacturing production workers: $30.19, about $5,230 a month (September 2026) | NBS; Eurostat; US Department of Labor; Ballotpedia; BLS |
| Vehicle development time | About 24 months (new carmakers) | 40 to 50 months (established carmakers) | McKinsey, August 2025 | |
| R&D as a share of GDP | About 2.8% (2025) | 2.2% (European Union), 3.1% (Germany), 2.18% (France), in 2024 | 3.37% (2023, preliminary), about 3.39% (2024 estimate) | NBS; Eurostat; SIES; NSF NCSES |
Chinese public support in detail
CSIS, an American research centre, estimates that China spends at least 1.73% of its GDP on industrial policy, around $248 billion in 2019. Even with this conservative estimate, it spends far more than the seven other economies compared, including France, Germany and the United States43.
The Kiel Institute, which advises the German government, has quantified the case of BYD: at least €3.4 billion in direct subsidies between 2018 and 2022, or 1.1% of its revenue in 2020 and 3.5% in 2022. Almost all listed Chinese companies received subsidies in 202244.
Part of this support does not appear on any subsidy line.
- Industrial land sold by local authorities below market price alone represented 0.21% of GDP in 2019, according to CSIS45.
- Electric vehicle buyers benefited from bonuses, then from a purchase tax exemption of around $4,200 per vehicle in 2024 and 2025, according to Kiel46.
- Subsidised battery makers sell more cheaply to their carmaker customers, which indirectly benefits BYD47.
The European Commission concluded its anti-subsidy investigation in October 2024. It imposed countervailing duties for five years of 17.0% on BYD, 18.8% on Geely and 35.3% on SAIC48. Beijing disputes these findings and has taken the case to the WTO49.
The method in detail
The five steps summarised in chapter 4, with the programme calendar.
Everything starts from the strategic vision
In China, everything starts from a direction: the five-year plan, BYD's mission, XPENG's roadmap to 2030. Without a direction, a transformation becomes a collection of proofs of concept that never scale.
- Where the company wants to win.
- What manufacturing must make possible in lead times, costs, quality and flexibility.
- Three to five priorities, quantified in P&L impact.
This is a business vision, in which data and AI are means. Improvements from the shop floor should be encouraged, but major transformations must serve this vision, with enough impact to justify their cost. Once approved by the executive committee, the vision filters every choice. The CEO carries it personally, with short messages.
The diagnostic measures the gap between the vision and the shop floor
Over four to eight weeks, the diagnostic tests the vision against reality, through interviews at headquarters and on sites, field visits and a review of existing data. Each finding is cross-checked against several sources, then grouped by cross-cutting problem, whatever the site.
The diagnostic results in around ten proven findings and a map of candidate use cases.
From the readout, a Notre-Dame plan lightens the rules
At the readout, the executive committee chooses three pilots using an impact-effort matrix, filtered by the vision. It favours frequent, replicable and measurable cases, including one where the problem is acute and the will to act is strong. For each pilot, it sets down in writing the sponsor, the budget, the governance and the criteria for deploying or stopping it within the quarter, as well as the exemptions from group rules that slow down iteration.
In detail
This decision addresses a common problem. In the West, most pilots succeed, then plateau. The block can come from a purchasing cycle, employee representatives consulted too late, middle managers whose objectives have not changed, an IT roadmap booked for eighteen months, or a CEO who prefers launching another pilot to making a call. Each of these brakes is reasonable. Together, they turn a six-month technology into a three-year transformation, deployed two generations late.
Modelled on the special law that allowed the cathedral to be rebuilt in just over five years. Within a defined scope and timeframe, the transformation is exempted from some administrative and IT constraints, and its human obstacles are dealt with upfront. In practice:
- shorter approvals;
- simplified purchasing below a set threshold;
- a secure sandbox to test tools without waiting for their integration into the information system;
- employee representatives involved from the launch;
- middle-management objectives rewritten to serve the transformation.
Acceleration comes from lighter rules, with no extra effort asked of teams.
Six months of pilots: speed of iteration before perfection
“Forgetting the lesson of failure will make you fail again.”
Slogan displayed at one of the sites visited
The first change is cultural: you have to give up "right first time" during build phases. The companies we visited launch early and correct as they go. Too often we wait for a perfect project, which arrives late, or never.
The three pilots therefore run in parallel, and one of them may fail. That is expected, as long as it fails fast and teaches a lesson. Each pilot is deployed or stopped within the quarter, and none is left to drift. Each follows the path observed at XPENG:
First wins must be visible and measurable within the first hundred days. For example, an improvement proposed and deployed by operators, then rewarded publicly, or a key supplier supported by an improvement programme. Each win goes into a gains register, quantified against the starting point.
Short rituals build a culture of speed
Four rituals, inspired by what we saw, take hold within a few weeks.
A one-page weekly governance tracks progress, risks and decisions. The employment question is dealt with from the start: a skills and redeployment plan says what people do when you automate.
In the sixth month, you decide whether to scale up
The month-six review draws on the gains register, certified by management control. Each pilot has already been deployed or stopped within its quarter. The review decides on phased replication, documents what failed and updates the vision for the next three years.
The timeline
| Step | Duration | Deliverable |
|---|---|---|
| Vision | Weeks 1 to 2 | Vision page approved by the executive committee, sponsor named |
| Diagnostic | Weeks 2 to 8 | Proven findings, state of the data, map of use cases |
| Readout | End of month 2 | Pilots chosen, Notre-Dame plan, decision criteria in writing |
| Pilots and rituals | Six months | First wins in the first hundred days, rituals in place, gains register |
| Programme review | Month six | Review of pilots deployed or stopped, phased replication plan |
43. CSIS, Red Ink, May 2022: conservative estimate, excluding instruments that cannot be quantified.
44. Kiel Institute, Foul Play?, April 2024, reported by Bloomberg.
45. CSIS, Red Ink, May 2022, p. 18.
46. Kiel Institute, April 2024, reported by InsideEVs.
47. Kiel Institute, April 2024, reported by Bloomberg.
48. European Commission, Implementing Regulation (EU) 2024/2754.
49. Request for consultations at the WTO, reported by Omnitrans.