--- title: "No dashboard on purpose: Marc Olivier Meunier of Astra | Skene" description: "Astra sells traceability to lawyers and tracks almost nothing about its own users, so the team keeps calling customers. Marc Olivier Meunier explains why." url: https://www.skene.ai/resources/the-question/marc-olivier-meunier doc_version: "c00117dbc7a2" last_updated: 2026-10-30 --- Resources / The Question / Issue 6 # Marc Olivier Meunier: why Astra runs without a dashboard Marc Olivier Meunier. COO and co-founder of Astra, a brand protection company. Issue 6 of The Question, published by A Founder's Friend on 18 August 2026. By [Nicolas Dolenc](https://afoundersfriend.substack.com), author of A Founder's Friend. Skene adapted this page from his published issue. Astra tracks one thing about its own users: who opened the app in the past few days. Not minutes spent and not actions taken. Marc Olivier Meunier says this is deliberate, and the series editor called his reason the most direct answer the series had given so far to what a young company should trust. **The short answer:** Trust lives in an evidence chain a person can check, so at ten months the team keeps its own analytics near zero and talks to customers instead. ## What Marc Olivier Meunier trusts, and for what | Evidence | Trusted for | Not trusted for | | --- | --- | --- | | Provenance and a score on every item Astra shows | What customers buy: answers they can check | Explaining how the models work, which stays private | | Proprietary models | Industries where they perform well | Others, where a general LLM bridges the gap | | Who opened the app recently | Knowing who is active | How they use the product | | Feedback sessions and calls | Where the product goes next | Tracking consumption the team cannot see | ## Traceability is the product Astra finds brand impersonation, counterfeit listings and fake shops for lawyers and brand protection teams, scores what it finds and drives takedowns. The team found deep dissatisfaction with existing vendors and identified opacity as one of its main drivers. So every item Astra shows carries where it was found and how it was scored. The method stays private; the answer is checkable. ## No dashboard, on purpose Marc protects the team from having a cheaper alternative to a conversation. If a dashboard existed, someone would check it instead of calling the customer, and with a handful of accounts and legal buyers who will say exactly what is wrong, the call is the better instrument. > From a tracking perspective, we only check who has been on the app in the past few days. Marc Olivier Meunier, in Issue 6 ## Trust in a model is per industry He is candid that the same model can be excellent in one industry and poor in another, so the system routes the weak cases through a general model. Trust here is granular and maintained, not declared once. > Sometimes we realize that our model is extremely effective for one industry and it can be really, really dumb for another industry. Marc Olivier Meunier, in Issue 6 ## The $250,000 nobody watched The one place missing tracking cost Astra money was spend, not behaviour. The company had $250,000 in Cloudflare startup credits. With 60 days left on them, it had used about $500, including an AI gateway that could have carried much of its model inference the whole time. That is the same failure as untracked product usage, applied to consumption. Marc's own conclusion is that they could have used a lot more of it. > in 10 months we’ve used about $500 of it. Marc Olivier Meunier, in Issue 6 ## What to do with this Deliberate absence is a reasonable choice at ten months, and it has an expiry date nobody announces. Decide now which signal will tell you to start measuring, whether that is account count, an agent shipping code faster than anyone documents it, or spend you cannot see. For seven checks drawn from across the series, see the [questions to ask before trusting a number](https://www.skene.ai/resources/the-question#before-you-trust-a-number). ## Frameworks this interview tests - [Customer health score](https://www.skene.ai/resources/frameworks/customer-health-score): A customer health score blends usage, support and sentiment to predict renewal risk. Usage carries the most weight, and nothing in the score checks the usage events. - [Churn prediction scorecard](https://www.skene.ai/resources/frameworks/churn-prediction): A churn prediction model flags at-risk accounts from leading indicators like falling usage. A broken event looks just like falling usage, and it trains the model wrong. ## Read next in The Question - [Dino Repo: one metric, five numbers, and a ten-minute fix](https://www.skene.ai/resources/the-question/dino-repo) - [Emad Ghorbaninia: five analytics tools or one five-minute call](https://www.skene.ai/resources/the-question/emad-ghorbaninia) - [Hamid Farmani: talk to people before you track anything](https://www.skene.ai/resources/the-question/hamid-farmani) - [The sponsor answers The Question: Teemu Kinos, cross-examined](https://www.skene.ai/resources/the-question/teemu-kinos-richard-makara) - [What all fourteen interviews add up to](https://www.skene.ai/resources/the-question) ## Source and disclosure [Read Issue 6 on A Founder's Friend](https://afoundersfriend.substack.com/p/aff-the-question-issue-6-making-decisions). Quotations on this page are copied from the published issue. Skene sponsored this research and commissioned it on a single question. A Founder's Friend ran the interviews, wrote them up and published them, and the series states that Skene paid for the research and did not direct the questions, the analysis or the conclusions. Each of the thirteen interviews was reviewed and signed off by the person in it, named or anonymous by their choice. Appearing in the series is not an endorsement of Skene, and nobody here is presented as a Skene customer. ## Sitemap - [Markdown sitemap](https://www.skene.ai/sitemap.md): every page on this site, with its markdown counterpart. - [Index for language models](https://www.skene.ai/llms.txt): the same set as a link-dense index. - [Full corpus](https://www.skene.ai/llms-full.txt): reference content concatenated as plain text.