Inside the decision product

Sample Decision Brief

From Fit Check to a First China Experiment

RecommendationPROCEED

Begin with Beijing E-Town / Yizhuang validation.

China functional fit: HIGH
Beijing fit: HIGH
Immediate entity requirement: LOW
Manufacturing dependency: MEDIUM now; potentially HIGH later

The project and the decision

A Belgian robotics company headquartered in Leuven has a 14-person engineering team, working mobile-manipulation prototypes and early customer trials outside China. Its robot navigates to a workstation, identifies an object and moves it into position. Controlled demonstrations go well. Performance deteriorates when lighting changes, objects arrive at unfamiliar angles, people interrupt the route or surfaces differ from the training environment. The next customer deployment needs dependable operation across that variation.

The team asks: “Should we establish a China company to access the robotics ecosystem?” Our recommendation for this illustrative project is to proceed with a focused Beijing validation route. China offers a strong functional fit, and Yizhuang is the first place to test that fit. An entity can follow once repeat work, local contracting or sustained engineering activity gives it a clear purpose.

First test whether Beijing can make the robot better. The initial commitment should buy evidence about reliability and iteration speed, with enough engineering access to act on what the tests reveal.

Turn the expansion question into an engineering brief

The limiting factor is the cycle between failure, diagnosis, modification and re-test. The team can already build a functioning robot; it struggles to expose enough different failures and resolve them quickly. A useful China engagement therefore needs varied physical conditions, people who can diagnose system behaviour, and a practical way to change components or assemblies between test runs.

We would ask for a non-confidential task description, representative failure categories, current intervention frequency and typical modification lead times. Those inputs let us separate perception errors from grasping, navigation, integration and mechanical problems. They also give prospective hosts a specific question: can you help this robot complete these tasks under these conditions, and support another test after we change it?

For this sample, the screening sequence below illustrates how a broad search becomes an actionable shortlist. The counts are demonstration figures. Three options warrant direct verification; Yizhuang is the recommended place to begin those conversations.

Illustrative screening · demonstration counts

18Potential routes
7Technically relevant
5Operating fit
3Worth direct verification
1Recommended first route

Why Yizhuang comes first

Beijing E-Town’s January 2026 account of the Beijing Humanoid Robot Innovation Center pilot verification platform describes prototype production, performance and process verification, assembly, testing and small-batch trial services. For this team, that creates a concrete prospect: investigate an assembly or component failure, modify the relevant part, and test the integrated machine again. The first technical enquiry would establish which equipment and procedures can support the team’s mobile-manipulator configuration. [1]

The August 2026 E-Town update describes E-Bot Hub’s shared laboratories, robot hardware, technical support, laboratory booking and supply-chain prototyping services. These capabilities could give a visiting engineering team a practical base for repeated tests and modifications. We would verify bring-your-own-robot access, the available engineering support, booking dates and the handling of test data. The same update describes the Jiyin Gongchang supply-chain platform for component selection, prototyping and production coordination; a specific drawing and quotation would establish turnaround for the required change. [2]

Yizhuang’s embodied-intelligence social experiment programme identifies settings including shops, hotels, pharmacies, hospitals and logistics warehouses. That range matters because the robot’s weakness emerges outside familiar demonstrations. A host conversation would select a controlled test setting that reproduces the relevant clutter, movement and object variation, then agree supervision and operating conditions. [3]

Our assessment is that this combination deserves the first verification effort. The comparison below keeps other routes available for the function they can best serve. Incubation and policy support can strengthen the eventual operation; the immediate engineering test determines where that operation should begin.

Sequence the work by the problem it solves
Route	Sequence	Purpose
Beijing E-Town / Yizhuang	First verification	Combine physical validation, engineering changes and repeat tests within one working visit.
Wider Beijing research ecosystem	Targeted parallel enquiry	Bring in a specialist if failure analysis points to perception, control or manipulation research; test the China–Belgium cooperation background against that specific need.
Jing-Jin-Ji manufacturing network	Prepare after validation	Compare fabrication, assembly and quality-control options once the design and first batch are defined.
General incubator package	Later operating support	Useful workspace and business support when the project needs a recurring local base.
Incentive-led location search	After technical fit	Assess financial support against the selected work programme and actual operating commitments.

What the Belgian home base changes

The Leuven base changes the sequence in four practical ways. Mobility matters before the first visit; tax and investment arrangements become useful as the operating commitment grows; research cooperation can be investigated alongside validation. Each layer therefore enters at a different decision point, while Yizhuang remains the first capability to verify.

For engineers holding Belgian ordinary passports, China’s current unilateral policy permits business and exchange visits of up to 30 days through 31 December 2026. This lowers travel-administration friction around the proposed ten-working-day visit. The company’s Belgian headquarters alone does not determine each traveller’s eligibility: confirm passports and that the planned activities fall within the published scope before booking. [4]

Recurring work brings the China–Belgium tax treaty into the operating review. Article 5 covers fixed-place permanent establishments and services for the same or connected project exceeding 183 days in any twelve-month period. Article 12 caps qualifying royalties at 7% of gross payments where the beneficial owner is resident in the other state; its definition reaches patents, know-how, designs, trademarks and equipment relevant to robotics. Article 10 provides a 5% dividend ceiling for a qualifying corporate beneficial owner, excluding partnerships, directly holding at least 25%, and 10% in other treaty cases. Its original uninterrupted twelve-month holding condition must be read with the MLI’s 365-day period including the payment date. These are conditional source-tax ceilings. [5] The current MLI rules also include a principal-purpose test for treaty benefits. [8]

A later qualifying Belgian investment would bring in the China–Belgium-Luxembourg Economic Union investment agreement, signed on 6 June 2005 and effective from 1 December 2009. MOFCOM lists it as currently effective; it replaced the 1984 agreement. Its investment-treatment, protection, transfer and dispute provisions belong in the review when the parent commits capital to a lasting China operation. The proposed visit should first establish whether such a commitment serves the engineering programme. [6]

A research problem could open a parallel enquiry. China and Belgium signed their economic, industrial, scientific and technological cooperation agreement in 1979 and established a joint-commission mechanism. The twentieth federal S&T joint commission met in Beijing in November 2023, alongside a history of innovation dialogues. If validation exposes a perception, manipulation or control problem requiring research collaboration, this gives the wider Beijing research route an institutional background to investigate. A specific partner, collaboration channel and project fit would still need to be established; no open grant or company eligibility is assumed here. [7]

When each Belgian-specific layer enters the decision
Belgian-specific layer	Relevance	Decision effect
30-day visa-free business / exchange entry	Immediate	Makes a qualifying short validation visit easier to activate.
China–Belgium tax treaty	Later, if operations deepen	Review PE, IP licensing, intercompany payments and eventual dividends.
China–BLEU investment treaty	If a qualifying investment follows	Review the framework around the Belgian parent’s China investment.
China–Belgium S&T cooperation	Potential parallel route	Investigate research collaboration if the engineering evidence warrants it.

A first Beijing experiment with a decision at the end

Plan a ten-working-day validation visit after a host confirms the test environment, equipment access and engineering support. Before travel, the team records a baseline on its current task set and agrees a trial budget covering host services, modifications and logistics. The engineering lead owns robot operation and changes; the host supplies the agreed facilities; Meridian coordinates the working questions, arrangements and decision record.

For eligible Belgian passport holders, the current business/exchange entry policy supports activating this short visit once the host and travel conditions are confirmed. Robot transport, the host schedule and the agreed test activities remain part of preparation. [4]

During the first two days, reproduce the main failures and agree how interventions will be counted. Days three to five introduce unfamiliar layouts, object positions, lighting and surface conditions while keeping a repeatable task set. The team and host then diagnose the highest-impact failures, make selected mechanical or integration changes, and repeat the tests during the remaining days. Reserve a held-out set of conditions so improvements are tested beyond the settings used for tuning.

Illustrative continuation criteria are shown below. The team and host would set the final thresholds before work begins, using the actual baseline and customer requirement.

Reliability: improve successful completion by at least 10 percentage points over the matched baseline, across at least 100 documented attempts per configuration; record interventions and any safety regression separately.
Iteration speed: complete two documented modification-and-retest cycles, with at least one hardware change ready for re-test within three working days.
Distinctive access: demonstrate useful test conditions or engineering support that the team would struggle to reproduce at its current base.
Next-stage economics: obtain a named engineering lead and an itemised route for the next batch, with lead times and costs that compare favourably with the team’s existing alternative.

Continue if the results establish a material engineering advantage and a workable next collaboration. If performance barely changes, retain the failure analysis and modification record, then redirect effort to the unresolved subsystem or another host. Either outcome gives the team a firmer basis for its next investment.

Revisit incorporation when repeated use of the Beijing engineering environment, local hiring or recurring contracts makes a China entity operationally useful. The first visit should establish that dependency and clarify the structure needed to sustain it. At that point, map the Belgian parent’s operating presence and PE exposure, ownership and licensing of IP or know-how, intercompany payments and the investment framework against the actual recurring work. Those questions can then inform the entity decision.

The work Meridian does, and the files you receive

Public knowledge helps a team understand the available landscape. The free Fit Check provides an automated structural screen; an email conversation can then define a specific engagement. In this example, the paid work would translate the engineering problem into counterpart-facing questions, verify the most relevant capabilities, prepare an agreed project summary, coordinate scoped conversations and turn the answers into a recommended experiment. Successful validation could lead to a separately agreed manufacturing workstream.

The finished package should let the engineering lead explain the recommendation internally and assign the next actions. Depending on the agreed written scope, an initial engagement may produce a combination of the outputs below. The indicative US$1,000–1,500 range covers a defined initial assignment; the full validation visit, extensive outreach, physical testing and third-party engineering costs are priced according to the work agreed.

Decision Brief: A readable PDF with the recommendation, alternatives, evidence and conditions for continuing. In this example: proceed with Beijing validation and defer incorporation.
Capability Shortlist: A comparison sheet linking each surviving option to the engineering task, its public evidence and the next access question.
Counterpart Record: For contacts included in scope: organisation, relevant function, conversation date, confirmed support and the person responsible for the next step.
Verification Log: A dated record separating published capability, direct answers and unresolved questions, so the team can see what the recommendation rests on.
Recommended Experiment: A working test plan with inputs, tasks, host responsibilities, budget assumptions, schedule and continuation criteria.
Next-Action Handover: An owner-and-deadline list the engineering lead can use immediately: who prepares the robot, who confirms the host, and what triggers the next decision.

Evidence and the next conversation

The register below records the public evidence checked on 8 September 2026. It supports the capability assessment; the experiment design and route recommendation are our analysis of the fictional project. When a route depends on a fact that public sources or Meridian’s internal capability and policy intelligence cannot resolve, we verify it directly with the relevant office, platform, institution or operating counterpart. Typical questions include access, eligibility, timing, equipment acceptance, data rules and the correct operating window. This fictional sample stops before that direct-verification stage.

Bring the task your system must perform, where it currently fails and the milestone you need to reach. We can use that starting point to define the first useful piece of work and the decision it should produce.

Discuss your next experiment →Start the Free Fit Check →

How We Work →