How to structure the Data & Analytics team? Where should Data be positioned in the organization? Which profiles should be recruited first?

To get answers to these questions faced by many CDOs, we interviewed Olivier Monnier, Chief Data Officer at Matmut, an insurance group employing 6,300 people with a turnover of 2.9 billion euros. 

Olivier, thank you for receiving us, can you please introduce yourself?

I am Olivier Monnier, Group Chief Data Officer at Matmut for the last two years. Before that, I worked for Roquette, HSBC and Scor in very different contexts.

Can you tell us how you became Chief Data Officer at Matmut?

There was a desire from Matmut COMEX to industrialize the Data & Analytics initiatives within the group. In 2015, the group created a  “Datalab” within the IT department, which launched many small Data projects. 

The Wanted wanted to scale up and move beyond experimentation. The goal was to take full advantage of new technologies such as AI to win customers and increase operational efficiency. 

Where is Data positioned within Matmut?

As I mentioned before my arrival, there was already a Datalab attached to the IT department. However, we decided to create a fully-fledged Data Department reporting directly to the COMEX. The idea was to create a bridge between the business and the IT department, because they have different rhythms.

How did you structure your team?

I was alone when I started  but two years later there are 15 people in my team.  

I have structured my team into 4 different areas:

– Data Analytics 

– Data Science 

– Data Engineering 

– Data Management

 

I decided to staff two roles in priority : Data Analysts and Data Engineers. 

The Data Analysts were essential for me because they are the bridge with the business. In terms of profiles, I hired data experts who didn’t really want to work on technical issues anymore but rather spend time understanding the business issues. 

Then I spent a lot of time staffing Data Engineers whose skills are increasingly rare on the market, but with patience and determination, I managed to hire them. 

How are you dealing with the talent crisis?

Talent management is a priority for us. I don’t have a miracle recipe, but I have adopted the following strategy: I don’t try to staff the best talents but I try to staff a complementary team. You also have to be ready to help profiles that are not yet fully mature to become more competent, patience is essential. The Matumut is in Rouen, so we have an additional difficulty. 

That’s why we worked a lot with HR, we spent time with recruiters to train them to  Data & Analytics. We created a recruitment process that includes technical tests to qualify our candidate earlier & better and improve the recruitment experience.

How did you prioritize the first use cases to generate value?

First, I have a basic principle… I don’t do PoC (Proof of Concept). 

Seven or eight years ago, we used to do POCs to evaluate whether a data project could deliver value or not. Today, this point is no longer to be proven. The answer is obviously yes. 

I focus only on operational & scalable usecases and before to start working, I need to validate 3 points: 

1) Why this usecase? 

2) What value could be generated? (I’m talking about value, not ROI ). 

3) What is the risk of not doing it? 

 

If the business is not clear on these three questions, we do not take the usecase. 

Then, at the Data team level, prioritize our projects, we set up a methodology based on 5 questions: 

1) Can we use the data (Reliability, Accessibility, Diversity)

2) What is the pre-processing effort? 

3) How complex are the algorithms to be created? 

4) What is the visibility of the project within the company? 

5) What is the impact on the business? 

 

We give a score to each of these points. At the end, the projects with the highest score are prioritized. 

How do you secure sponsorship? 

Sponsorship should be done at two levels. 

You have to work and explain your vision to the COMEX and also secure the relationship with the business leaders. 

My COMEX was convinced that Data & Analytics were very important for our business. 

Nevertheless the Business Leaders were not 100% convinced and not really willing to change their habits.  We had to get them involved and not let them out this data transformation. To do this, we proceeded as follows: first we met with the business managers, identified the potential  usecases with them, and then introduced the summary of our meetings to the COMEX. 

At this point, it is the COMEX who decides which usecase should be prioritized. This methodology allowed us to involve all the stakeholders.

What about your relationship with the IT department?

The IT department is involved everywhere, it is a strategic partner. I spent a lot of time with my CIO to define the roles & responsibilities of everyone.

What are the responsibilities of the Data Management team?

The Data Management team focuses on three major issues for us:

– Data quality 

– Data Catalog 

– The definition of Data Owners, up to writing a mission statement

In terms of technology stack, can you tell us about tools that have helped you in your role?

There is a tool that I naturally think of, because it has boosted us and changed the way CDOs think, that’s Datarobot. This type of tool is changing the way data scientists work. Thanks to this technology, the way data scientists code becomes secondary because the software can create good quality algorithms. 

In terms of recruitment this has a huge impact, 4 or 5 years ago we were looking for candidated who were the best in coding. Today, thanks to Auto Machine Learning, I’m looking first for profiles able to understand the business. 

The other advantage of this tool is that it takes into account our own Python codes developed outside the software.

In terms of governance within the Data team, which mode did you choose?

I had a very centralized management at the beginning to make my team mature, but the medium term project is to decentralize my team to go towards Data Mesh, which is an objective for us because it is a model I believe in.

What do you think of Hubadviser ?

I recommend Hubadviser because there is nothing better than connecting with peers to help you take a step back from your own situation. Getting objective, independent and operational insight has a lot of value to me.

Cela pourrait vous intéresser

Data Governance : Experience Feedback

By Isabel Gomez Garcia de Soria

Data & Analytics Team Structure : Experience Feedback

By Olivier Monnier

From Data Lab to Data Office, the journey of a Data Scaler

By Maxime Havez

Digital Factory: Feedback from Serge Yoccoz

By Serge Yoccoz