A deeptech project often starts with a meeting: a researcher who holds a technology, and someone who knows how to sell it. They get along, they find they complement each other, and they go straight to the founders' agreement (pacte d'associés), or worse, to nothing at all, working together "to see how it goes". Bpifrance publishes a template document to fill exactly that gap: a memorandum of understanding for an "engagement period". The word raises a smile. Yet it names the most poorly managed phase of starting a company: the one where people become partners without having tested it.
What an untested partnership breaks
We tend to assume a startup dies from a missing market or a technology that doesn't hold up. The data from Noam Wasserman, the Harvard professor who followed some 10,000 founders for The Founder's Dilemmas, says otherwise: most failures of high-potential startups come down to people problems, first among them tensions between cofounders.
65% of failures of high-potential startups come from people problems, led by tensions between cofounders, against 35% from product, market or management problems (Wasserman, reported by CNN Money, 2014).
These conflicts aren't born when they break out. They are there from day one, in the form of questions no one ever asked: how much time does each person really put in? Who owns what was invented? What happens if one of you leaves after four months? While things go well, nobody wants to ask them. When things go badly, it's too late to answer them calmly.
The principle: test through work, not through talk
The value of the Bpifrance template rests on a simple idea. You don't validate a partnership by talking about it; you validate it by producing together, over a bounded period, deliverables defined in advance. The document proposes setting:
- a test period: the template recommends not exceeding six months if the newcomer isn't paid;
- each person's time commitment, written down in black and white: full time, part time, evenings;
- three to five workstreams (strategy, the relationship between partners, day-to-day operations), each with a deliverable, an owner and a success criterion;
- a follow-up rhythm between the team members.
One detail of the template is worth pointing out: it advises against assigning titles, CEO or CTO, during this period. Official roles freeze a collaboration that hasn't found its shape yet. They get settled in the founders' agreement, after a few months of real work.
What this indicates — A partnership is proven like a product: through dated deliverables and a success criterion set beforehand, not through the enthusiasm of the moment. An i-Lab application written by two people says more than an evening spent discussing the vision.
The two blind spots that cost a lot
The document insists on two subjects founders almost always postpone, and which are precisely the ones that turn into disputes.
Intellectual property. When the technology comes from a laboratory, it belongs to the laboratory, and exploiting it requires a license negotiated with the technology transfer office (a SATT or equivalent). The template suggests a clause that attaches everything invented during the engagement period to the technology, not to the person, so that a partner who leaves doesn't walk away with part of the value.
The legal status of the newcomer. This is the point the document calls absolutely key. Working on a project without any contract, neither as a corporate officer nor as an employee, exposes the future company to having that work requalified. The template cites the case of a project supported by the Incub'Alliance incubator: a prospective partner who left without a contract asserted his work, emails in hand, and the startup had to pay him compensation.
"We'll sort out the status later" is the most expensive sentence of the engagement period. Unstructured work doesn't disappear when the partner leaves: it comes back as a claim.
The answer Incub'Alliance came up with is instructive: a partnership with Pôle Emploi (the French public employment service, renamed France Travail on January 1, 2024) that lets a jobseeker join the startup as part of a training internship. The status is regularized, and the agreement provides that no claim will be made on the intellectual property produced during the internship.
What we take from it at Orogen
This document isn't a contract, and Bpifrance repeats it on every page. That is its strength: it forces you to write down what you wouldn't have said, without the weight of a legal instrument. Faced with a project that rests on a team still taking shape, this is the grid we propose, with three added questions:
- If one of you leaves in four months, what do they take and what do they leave? The answer should fit in two lines. If it's vague, the intellectual property isn't settled.
- What deliverable, by what date, will prove you can produce together? Not an intention: an object. A funding application submitted, ten customer interviews written up, a demonstrator.
- What, at the end of the period, would make you say no? A test period without a failure criterion isn't a test, it's a delay.
The engagement period isn't meant to guarantee the partnership will work. It's meant to make sure a partnership that doesn't work ends in month six, cleanly, rather than in month thirty, in front of a lawyer.
Sources
- Bpifrance, Parcours de fiançailles — Cadrage de la collaboration entre associés, memorandum of understanding template (non-contractual document): test period, commitments, workstreams, masterplan, IP clauses, Incub'Alliance – Pôle Emploi case — Bpifrance PDF.
- Noam Wasserman, The Founder's Dilemmas, Princeton University Press, 2012 — press.princeton.edu.
- 65% people problems / 35% product and market split: CNN Money, February 24, 2014 — money.cnn.com.
- Pôle Emploi renamed France Travail on January 1, 2024: Law No. 2023-1196 of December 18, 2023 on full employment.
- On proof through deliverables rather than intention: see the resource The proof that a project exists.