Our path to implement a product approach to data (1/5)

Why this article?

Last year, we made a bold reorganization in the data team and created a new team of Product Managers for Data.

Wait… what is a Data Product? Simply said, it’s a product that facilitates an end goal through the use of data. It’s designed to solve specific problems, enable decision-making, or enhance user experiences by utilizing data as a core component. Data products can take various forms: dashboards, tables, files, data APIs.

There are already plenty of great articles about data products but not a lot on what it means for Data teams to switch from a project-approach to a product-approach.

This is the challenge I’ve been given one year ago.

I wanted to share our experience of a difficult, but necessary and rewarding transition.

Do you know Doctolib?

Doctolib is a European tech company with 2 missions:

We improve the daily life of health professionals.

We help people to be healthier.

We support health professionals with an innovative software suite, co-designed with them. Doctolib streamlines workflows and delivers advanced clinical solutions, such as electronic health records, diagnostic support, and prescription management. It also includes Patient Management Solutions for scheduling, teleconsultation, and practice efficiency.

For patients, Doctolib serves as a trusted health companion, helping them access care simply, securely, and proactively. Concretely, we facilitate care management through easier and faster access to care (through a medical provider directory, messaging or online booking) and provide tools to manage health (health records, health tracking, medical content and preventive care).

We definitely operate at scale. The 2,900 Doctolibers serve 80 Million patients and 400 000 health professionals across Europe. To give you a glimpse of the activity of our users: every month, 19 Million appointments are booked, 9 million documents are shared. Another example: we launched a consultation assistant that takes notes for health professionals during a consultation, summarizes them and organizes the information in the patient file. After 7 months, it’s been used for 3 Million consultations.

How it started and how we reorganized the teams

Back in 2023, I was leading one of the 3 large teams of data analysts at Doctolib. We were integrated within business and product teams, doing a little bit of everything: reporting, ad-hoc analyses to support decision making, and sometimes complex modeling.

This organisation pushed us to develop great product and business knowledge to be more efficient and bring more value everyday. We had become great partners with other teams.

We identified 3 main problems:

• Technical debt was high: The priority had often been given to lower delivery-time so that Doctolibers get faster access to insights. This resulted in a lack of governance (data quality, inconsistent KPI definitions, multiplicity of dashboards…), foundations that are very hard to maintain and costs that were out of control.
• Access to information was not easy: We built a lot of assets (datasets, KPIs, dashboards) that were addressing user needs but we did not treat them as a product. There were overlaps, unused assets, and we did not have the proper proactive maintenance routines.
• Our Data Platform was not fit to support future AI ambitions: The platform had been developed to support internal reporting needs. It was not mature enough to support AI use cases in the product and external reporting in terms of scale and reliability.

There were bright spots though. When we finalized the revamp of one of the major datamarts, every data user started to grasp what “easy access” meant. A simple question like “how many meetings have been planned in March and occurred in April for team North” that took 2 days before, was answered in autonomy by non-data users in 3 minutes. On AI, we actually had a machine learning platform already, but it was disjoined from the rest. We wanted to replicate this at scale.

Late 2023, we decided to make bold changes in the Data organization to:

1. Deliver more AI products, at scale
2. Enable Doctolibers to do a better job through self-served data products

We broke up the data analyst teams and reassigned everyone to new roles:

• Analytics Engineers: Build and maintain data products, from ingestion to exposure
• Data Scientists: Power products with AI
• Data and ML Platform: Build & operate the data platform
• Data Governance: Define, share and track data management policies
• Data Product Managers: Define and prioritize data products that maximize value for our users

To make our organization the most efficient, we segmented Analytics Engineering and Data Science teams based on Doctolib organization and positioned Data Products PMs in every sub-teams. These Data Product teams will interact with our Data and ML Platform team, organized by components (Data collection, Data Engineering Platform, ML Platform and Tools). And here again, we have positioned Data Platform PMs as the first points of contact.

When I accepted to take the lead of the Data Product Managers my challenge was to implement a product approach within the Data teams, to generate more ROI, more scalability, more user-centricity, more impact.

What we did and what we learned

This first year has been a year full of challenges and learnings. There is a lot to talk about, so I’ve dedicated an article to each of the topics below:

1. We rebuilt teams entirely with new roles and started by defining our vision that will drive us for the upcoming years (🔗 link to article)
2. We transformed our ways of working around the product mindset: we now leverage OKRs (Objectives and Key-Results) for prioritization and implemented new processes for discovery, delivery and operations.(🔗link to article)
3. We launched a project to completely rebuild the data platform to reach our AI ambitions and drastically improve the engineer experience. (🔗 link to article)

What’s ahead of us?

Yes, in 2024 we made significant progress. And yes, in 2025 we still have a lot of challenges to address.

We will continue our efforts to improve Doctolibers’ ability to 1. leverage data for insights and 2. to build better products, leveraging AI; both in autonomy and at scale.

We will onboard more and more people on our data products and data platform. It will come with a strengthened and federated data governance to make sure everyone contributes to the data value chain.

And obviously, we will keep developing our product approach to data.

Big shoutout to the teams for embracing the journey and for all the accomplishments and learnings that made these past months exciting and special.