A sensor works, a robot delivers a successful demonstration, a material passes its initial tests. At a VivaTech stand, collaboration seems almost assured. On the factory floor, everything is just beginning: qualification, integration, procurement, maintenance. For an industrial startup, the danger is not just technical failure. It is also running a successful trial that leads nowhere. In September 2026, this question deserves to be asked behind the partnership announcements: who turns proof into an order, with what budget and on what timeline?
The trade show showcases connections, rarely the path to scale
VivaTech, launched in Paris in 2016 by Publicis Groupe and Groupe Les Echos, has made connecting major corporations with startups one of its hallmarks. Demonstrations, innovation challenges, business meetings: the format provides access to contacts who would otherwise be difficult to reach. For a company developing a robotics component or a decarbonization technology, that visibility can cut months off the search for customers.
But a trade show does not operate on the same timeline as industrial investment. An innovation department may want to explore a technology without a factory having decided to buy it. A site manager may be convinced but lack the budget. A procurement manager may become involved only after testing and discover a supplier still unable to meet all their requirements.
Trends observed through 2024 shed light on this tension: the rise of reindustrialization policies, support for deep tech, and the search for energy and environmental solutions, alongside startup funding that became more selective after the euphoria of 2021. The analysis offered for September 2026 is forward-looking: it assumes neither specific announcements at the 2026 event nor an already established commercial track record.
A successful pilot is not yet a product ready for purchase
A pilot serves a legitimate purpose: reducing uncertainty. Before modifying a production line, a manufacturer must verify a solution’s safety, reliability and compatibility with its operating environment. A maintenance algorithm tested on prepared data may struggle with aging sensors. A material that shows promise in the laboratory must retain its properties when manufacturing volumes and conditions change.
The trap emerges when the initial question remains vague. “Does it work?” is not enough. Other questions must also be asked: does it work for long enough, at what total cost, with how much human intervention and for what operational gain? Without agreed criteria, a technical success may be followed by another request for testing, then another.
Imagine a startup whose vision system correctly detects defects on a production line. The factory values the results but now requires a connection to its IT systems, continuous operation and rapid support. None of this is unreasonable. Yet if those conditions were not anticipated, the pilot has validated a demonstration, not paved the way for a purchase. The difference comes at a cost in development work and cash flow.
Two organizations, two clocks
For a major corporation, a few extra months may fit within a normal budget cycle. For a startup, they mean salaries, capital tied up in components and a funding round to prepare. The imbalance is particularly pronounced in industry: building a prototype, sending technicians to a site and booking machine time cost more than providing access to software.
Public funding can support this phase. In France, Bpifrance programs, the Deeptech plan launched in 2019 and France 2030 have notably sought to support innovation and industrialization. But financial assistance is no substitute for demand backed by purchasing power. It can fund part of the technological risk without resolving the absence of anyone responsible for the purchase on the customer’s side.
The problem does not necessarily imply bad faith. Internal objectives diverge: innovation teams seek to explore, production teams to avoid disruption, and procurement teams to secure supplies. If no one takes ownership of the final decision, everyone can fulfill their role while the project remains stalled. The partnership then has plenty of supporters but no real buyer.
Customization: growth that is not really growth
A prestigious first customer opens doors. It can also consume the entire company’s resources. The customer requests a specific interface, a mechanical adjustment, then a particular reporting protocol. The team agrees to preserve the relationship. Gradually, it builds a solution that is difficult to sell elsewhere.
This work is not necessarily wasted: early customers often help define the product. The dividing line is reusability. A modification useful across several sites enhances the offering. A succession of exclusive requirements turns the startup into a dependent engineering consultancy, without always ensuring it is paid accordingly.
Intellectual property then becomes decisive. Who owns the improvements? Can the company reuse its methods and components? Does it have the right to name the customer as a commercial reference? Broad exclusivity, granted in exchange for a small trial contract, can close off more markets than it opens.
Prepare for the order before launching the trial
The right antidote is not to eliminate pilots but to structure them as a step in the decision-making process. That means bringing together, early on, the people who will operate, secure and buy the solution. A trial contract does not necessarily guarantee an order; it should at least make the potential path to one explicit.
- Define success: a few measurable criteria linked to the industrial need, with an agreed validation method.
- Identify the decision-maker: someone able to take responsibility for the deployment budget, not just the trial budget.
- Cost out what comes next: installation, training, maintenance, cybersecurity and operating costs must be part of the discussion.
- Set an exit point: a decision date, conditions for extensions and payment for additional work.
A paid pilot is a useful signal, but does not by itself prove that a market exists. More revealing indicators are the involvement of operational teams, access to the necessary data and work begun on the future contract. Conversely, a free trial with no timetable or procurement contact should be regarded as a risky sales expense.
Measure orders, not just collaborations
For trade shows, incubators and investors alike, the number of partnerships announced remains an incomplete indicator. Their subsequent progress should be tracked: recurring contracts, deployments across multiple sites, decision timelines and the proportion of development work that can be reused. This information is less spectacular than a signing ceremony on stage, but tells us more about how an industrial company is built.
What now? Looking ahead to future editions of VivaTech, the maturity of collaboration could be measured by the ability to say yes more quickly, but also to say no. A refusal backed by clear reasons frees up a team; a perpetually extended trial ties it down. For industrial startups, the most valuable partnership will not necessarily be the most visible: it will be the one whose deployment conditions were discussed before the prototype was even plugged in.
