Cap Clarté: Testing a Service Business with Lean Startup

Cap Clarté: testing a service business in six months, then choosing to stop

Alongside my full-time role at CPW, I used my own innovation method on myself: could a practical, organisational support service for people going through separation in French-speaking Switzerland find its market? I set out the hypotheses, built cheap and fast, measured, and made a clean decision.

My role

Founder: strategy, offer, content, website, operations

Method

Lean Startup applied to my own venture, AI-assisted operations

Timeline

Spring 2026 → stop decision, 26 Sept 2026

The idea: a gap between the lawyer and the therapist

Separation and divorce bring a wave of practical work: understanding the options, organising logistics, preparing for the lawyer, maintenance contributions, a parenting plan, telling the children. Lawyers handle the legal side and therapists the emotional side. My hypothesis was that a practical, organisational layer was missing in between, and that people in French-speaking Switzerland would value it.

I treated it exactly like an FMCG concept: a set of explicit assumptions to test, a minimum viable version, and success criteria decided in advance.

How I built it: cheap, fast, and designed to learn

Everything was built to be tested quickly and at low cost, alongside a full-time job.

A real offer, not a landing page. Individual support in three levels (Basic, Plus, Premium), six practical workshops (understanding your options, organising logistics, preparing for your lawyer, maintenance contributions, communicating about the divorce, building a parenting plan), a free self-assessment and a resource library.
Low-cost stack. WordPress with free Elementor on shared hosting, a Google Business Profile, WhatsApp for first contact. No agency, no custom development.
Content as the demand test. Practical articles on the questions people actually search for (shared custody in Switzerland, separation vs. divorce, amicable divorce, mediation or lawyer, expatriate divorce) to see what drew real visitors.
AI-assisted operations. I designed the full client journey as a documented system: 18 standardised communications from first contact to closing, session reports, a satisfaction questionnaire. Generative AI helped me draft and run it, so a one-person service could run on evenings and weekends.
Measure from day one. Google Search Console and Google Analytics were connected at launch, with a monthly review.

Four hypotheses, four answers

Hypothesis 1

People in French-speaking Switzerland look for practical support during a separation.

How I tested it: Articles targeting real practical questions, measured in Search Console.
What I learned: Confirmed for information: one article on shared custody in Switzerland alone brought 61 clicks from 2,400+ impressions, 41% of all search clicks. Visitors who accepted analytics stayed about 4.5 minutes per session on average.

Hypothesis 2

They will search for a service like this one.

How I tested it: Tracking searches for the service itself (“accompagnement divorce” and similar).
What I learned: Not confirmed. In six months, “accompagnement divorce” appeared in search results 3 times. People search for answers (custody, maintenance, lawyers), not for this kind of service. A service without a search habit needs another channel, so I tested referrals: I built a trust network of lawyers, mediators and therapists. It was only starting to work when I stopped. One therapist told me he might have had a client for me, who was still hesitating.

Hypothesis 3

A structured offer (packages, workshops) makes the value easy to understand.

How I tested it: Pricing, workshop and self-assessment pages, with their visits and clicks.
What I learned: Visitors did explore the offer pages (pricing, workshops, individual support), mostly after arriving on an article. Interest in the content did not become purchases: no paying client in six months. I helped one person partially, free of charge, and three or four people introduced through friends never made the call. The step from reader to client needed a different trigger.

Hypothesis 4

One person can run it alongside a full-time job, with AI-assisted operations.

How I tested it: An 18-step documented client journey and AI-supported content and admin.
What I learned: Operations: yes, the documented, AI-assisted system kept the workload light. Growth was the harder part: a referral network takes time and presence, and that is where the trade-off with my core innovation work became clear.

The decision: a clean stop

On 26 September 2026, I chose to stop Cap Clarté and put my energy into my innovation work. It's the decision I describe in Pivot, Persevere or Stop: read the evidence honestly, decide, and don't let a project become a zombie that quietly drains time.

Stopping is not the same as failing. The experiment answered its questions in six months, at a very low cost, and left me with first-hand lessons that I now bring back to large-company innovation.

What I take back to corporate innovation

Information demand is not service demand. Strong traffic on helpful content says little about willingness to buy. Test the purchase, not just the interest.

A warm introduction is not a booking. Friends’ referrals felt like demand; none of them called. On a sensitive topic, the first step costs the client more than you think.

Check the search habit before choosing the channel. Three impressions in six months told me more than any survey could have.

AI makes a one-person operation possible, not a distribution channel. It compressed the work; it didn't create the referrals.

Decide the stop criteria early, then respect them. It's easier to coach teams through a stop once you've made one yourself.

“The best innovation decisions I've made were not the ones where I pushed the hardest. They were the ones where I read the data honestly and had the conversation nobody wanted to have.”Amandine Devergies