Chief Product Officer / Europe
Decisions that changed what our customers could see
- Paid subscriptions
- 1.58m
- Lisandre Audio, June 2026; 190,000 of them through operator bundles
- Average monthly churn
- 2.8%
- First half of 2026, from 3.4% in 2022
- Company revenue
- €268m
- 2025; my product, design and data budget is €29m
Decisions & consequences
01Replacing an unlimited plan with hours a listener can see
By autumn 2022 Lisandre's main plan, Illimité at €14.99 a month, had 870,000 subscribers. About 61,000 of them listened to more than forty hours of books a month. They were 7% of the plan but accounted for 34% of its royalty hours, and their royalties averaged about €23 a month each. Gross margin had fallen to 31%, and renewals of several large publisher contracts in 2023 were expected to raise our per-hour rates by around 6%.
The executive committee had two proposals in front of it. Our finance director, Mathis Orcel, proposed applying the plan's dormant fair-use clause quietly: after thirty hours in a month, new releases would load behind a short queue. It would have protected the margin within a quarter. I proposed retiring Illimité altogether in favour of a fifteen-hour plan at €11.99, a forty-hour plan at €17.99 and an openly priced unlimited plan at €24.99. Our head of growth, Inès Kerbrat, accepted the structure but argued in writing against launching in spring, our weakest season for renewals. Solène Varadier chose the visible plans and the spring date.
We gave up a share of our most devoted listeners, a season of growth and about €4.1 million in revenue from holding existing subscribers at their old price for twelve months. Before launch we tested the plan choice with 9,000 new subscribers in Belgium. Every existing subscriber received a letter naming the plan their own listening history suggested, unused hours rolled over for a month and nobody was cut off in the middle of a chapter.
About 13,000 of the heavy listeners cancelled within six months and 22,000 moved to the unlimited plan. Monthly churn rose from 3.4% to 3.9% for two quarters, then fell to 3.1% by the end of 2024. Average revenue per direct subscription rose from €13.20 to €14.60, and gross margin reached 38% in 2024.
Any limit now written into our terms must be visible in the app. I would change two things. We built the counter on hours of recording, the unit our royalties use, and listeners at faster speeds thought it was running fast; 1,900 of them complained in the first month. I should have tested the display with heavy listeners. And Inès was right about the season.
Taking publisher obligations out of the recommendation model
In late 2023 our recommendation team released a new ranking model for the home screen. Listening starts from the home screen rose 9%, and within a quarter independent publishers' share of home impressions fell from 15% to 8%. That share mattered: in 2020 Lisandre had signed a public charter with the Réseau Tilleul des éditeurs indépendants promising them at least 15% of home impressions across a year. Five of our fourteen rows also contained unlabelled visibility packages sold to publishers, worth €7.4 million in 2023.
Tobias Wendling, our head of recommendation science, proposed adding weights for publisher independence and for paid packages and tuning them until both commitments were met. It was a sound engineering answer. I opposed it because a weighted ranking would make three different promises impossible to tell apart, for the listener and for our own executives, and the trade-offs would be settled each quarter in a tuning file. Arnaud Pellequer, our commercial director, opposed my alternative more strongly than he opposed Tobias's, because labelled placements would sell for less. The argument took six weeks and went to the executive committee with both memos attached.
We adopted twelve rows. At most two carry sponsored placements, marked as promoted by the publisher. One is chosen weekly by our editors, who report to the content director. The rest are personalised, each title with a short line giving the main reason it appears, and the model is asked to optimise for relevance alone. We wrote the explanation of our recommendation parameters in plain French and Spanish and offered a home screen without personalisation.
Visibility revenue fell to €5.1 million in 2025. Home listening starts dipped 2% in the first quarter after the June 2024 launch, then finished the year 3% above the old baseline. Independent publishers took 16% of home impressions in 2025 and 13% of listening starts. About 4% of subscribers chose the non-personalised home screen, disproportionately those over sixty.
Our CSR committee now receives a quarterly account of the editorial row by publisher size and language. What I would change is the order of consultation. We showed publishers the label after choosing its wording, and some read promoted as advertising; we reworded it twice. I would also have given Arnaud's team a new product to sell, such as sponsored collections, before removing the old one.
A free podcast tier with limits written into it
In 2024 podcasts were available only inside Lisandre's paid app, while several larger services offered them free. Our cost of acquiring a paid subscriber had risen to €41, and research showed that many people who might pay for audiobooks first formed a listening habit through podcasts. The board approved a free tier funded by advertising, to launch in November 2024.
The disagreements were about what the free tier could contain. Mathis Orcel wanted audiobook excerpts of up to an hour inside it, with advertising, as a conversion tool. Our advertising sales lead wanted personalised targeting on by default, because contextual inventory sold for roughly 40% less. I argued for three limits: no audiobooks beyond samples publishers had approved, a cap on advertising minutes per hour of listening and personalised advertising only with the listener's consent. The chief executive accepted all three. We gave up advertising revenue and probably some conversions. The launch also slipped to March 2025, because I asked for two quarters of capacity for our older-listener work first.
My team designed conversion around moments of real interest, such as the end of a narrative series, and kept interruptions out of the listening itself. Creators who host with us receive a share of advertising on their episodes, paid monthly with a statement showing the impressions behind it. We measured movement downwards as carefully as movement upwards.
By June 2026 the free tier had 2.3 million monthly listeners and had brought 61,000 people to a paid plan. Advertising earned €9 million in the ten months of 2025. Some 46% of listeners consented to personalised advertising. About 6,000 paid subscribers who listened only to podcasts moved down to the free tier, roughly 0.4% of the paid base, which we report to the board beside the conversions.
The free tier now has its own written rules, reviewed by the board each year, about what it may never contain. What I would do differently is the creator share. We set it as one rate for every creator, and several small producers of narrative series, whose listeners converted most often, earned less than larger chat shows that converted almost nobody. A share linked partly to conversions would have been fairer and better for us.
Putting the bill before the offers in a mobile operator's app
When I joined Merovia Mobile in October 2016, its customer app had 1.1 million monthly users and a rating of 2.6 in both app stores. Research showed that most people opened it to check their bill or remaining data. The first thing they met was a carousel of offers, which brought in about €14 million a year of tariff upgrades and add-ons. Meanwhile Merovia's care centres handled about 2.6 million bill-related calls a year, at roughly €8 each.
I proposed in 2017 that the app open on the bill and data balance, explained line by line, with offers moved to a separate tab and shown only when a customer was eligible. I also proposed telling customers when a cheaper tariff would suit their usage better. Jolien Maertelaere, our marketing director, argued that the carousel funded much of the digital team and that the cheaper-plan prompt would cut revenue for no reason a customer would ever thank us for. Our chief commercial officer, Griet Hoorebeke, approved the bill-first design and limited the cheaper-plan prompt to a twelve-month pilot.
We gave up about €3.2 million a year of in-app sales, a fall of 23%. The cheaper-plan prompt moved 84,000 customers to lower tariffs in its first year, costing about €4.8 million of annual revenue. We worked with care agents to write the bill explanations in the words customers used on calls, and we rebuilt the contract-change journeys so that customers could change plans without phoning.
By 2020 bill-related calls had fallen 27%, saving about €5.6 million a year in care costs. Monthly app users reached 1.6 million and the rating rose to 4.3. Customers who had received a cheaper-plan prompt left Merovia at about half the rate of comparable customers who had not, and the board's customer and commercial committee asked for the prompt to become permanent.
Self-service became the default design for every contract change, with 85% of changes made without a call by 2020. What I would change is coordination with the retention team. For several months their agents were offering discounts to customers whom the app had just moved to a cheaper plan, and a few people received both. I should have brought retention into the pilot from the first week.