A day in the life of a data analyst in Criteo: meet Cecile Lefevre-Ardant.

By: CriteoLabs / 21 Nov 2017

We are often asked what kind of work Data Analysts & Data Scientists do at Criteo. Cécile has agreed to answer some questions, to give us an insight into her day to day work as a Data Analyst and her approach to new projects at Criteo.

Cécile Lefevre -Ardant- Analytics Lead Product, Criteo

What does the Analytics team work on?

The Criteo Analytics team has the privilege of working with almost all the other teams in Criteo. Some Data Analysts & Scientists are working closely with the commercial team to optimize campaigns to each client’s needs. Others are fine-tuning our internal marketplace to improve the overall performance. We even have a whole team dedicated to building awesome Data Science products used internally by everyone. And that is just scratching the surface of everything we do.

I work in the Product Analytics team, where we help new projects by extracting impactful and truthful insights from data. As part of the Product Team, we work closely with the Criteo R&D team.

And what does your usual day look like?

I participate in the R&D daily stand-up 10 minutes meetings, following closely how the project develops. As I manipulate data all day, I can often explain weird results that baffle developers and save them a lot of time. I can also make sure new releases that will impact my work are going well. It’s great to be closely involved in these processes as I can really see the impact of my results in the course of the company.

To the analyses side of my job. In a typical day, a Product Analyst works on a long-term major project that will answer a complex question. We take vague but critical problems such as “Is it worth investing more in real-time bidding developments?” and transform it into a complete analytical project that will give an actionable insight driving the roadmap. These often take several days to complete as it requires multiple models to be built (“Are we answering fast enough?” “Is there a correlation between size of the ad and performances?” “Is there significant inventory we are not addressing at the moment?”).

As I am currently working on Invalid Traffic Detection, I tend to have a bit more quick, short-term analyses than the average Product Analyst or Product Data Scientist. We are trying to identify patterns to help us detect internet traffic not resulting from genuine human interest (bot, forced clicks, etc.). I have a lot of ideas to test and I can immediately recommend shutting down some part of the traffic if it’s too suspicious. For instance, I found yesterday that a specific mobile app saw its traffic rise by 10000% in one hour along with terrible performances for our clients. It might not be fraudulent traffic, but it’s suspicious and costly enough that we can act immediately.

What kind of tools do you use?

The origin of all our data are the raw logs generated by Criteo applications. All Data Analysts & Data Scientists have access to a fantastic 140PB Hadoop cluster. We usually access it through Hive, a SQL-like software. The Analytics community is very much involved in sharing tips and proper usage on all of these amazing datasets. There is also an R&D team dedicated to making sure all of this juicy data is correctly transformed and aggregated for our usage. We can also build our own custom jobs to create new aggregated tables or send a copy to Vertica (our faster reporting infrastructure).

Analytics Team

To manipulate data, Data Analysts & Data Scientists can use what they prefer! We tend to favor R (we have a custom Criteo package), but you are free to use Python or Tableau instead or even Excel if you are really good at it. The only constraint we have is that your work should be understandable and re-usable as much as possible. We use Tableau as a reporting tool as its visualizations capabilities are just excellent!

What I really like is that we have a bunch of custom tools developed or forked at Criteo. For instance, we have a custom repository to showcase our best work called the Knowledge Repository. It’s a nifty combination of git and RMarkdown. This allows us to broadcast the finest analyses and models to the whole 200+ Analytics community of Criteo.

What do you enjoy the most about working at Criteo?

People are pretty cool and open-minded too. This is the only company I’ve worked for where you might hear three different languages during a single coffee break.

I appreciate how Criteo takes good care of its employees. The office is amazing as you can probably see from our corporate material. On my first day, I was setup with a desktop computer, plus a laptop AND I already had access to all applications relevant to my work. Also, I have the option to work from home when necessary.

Criteo really aims at having fair performances reviews. Your annual review is done by your manager, as well as your peers and is reviewed by another manager to avoid bias as much as possible.

My job is very focused on a specific slice of our business. However, I have the opportunity and I am encouraged to change the perimeter of my job every year or so. We also have an extensive process for sharing updates with all teams. I learn new things every day.

I get to seat frequently with the people I work closely with, such as the development team. Being agile with our team interactions allows me to have an understanding of what each person does and offer my assistance if needed The overall project moves forward much faster!  And it’s always more interesting to see the impact of your work in real-time.


  • CriteoLabs

    Our lovely Community Manager / Event Manager is updating you about what's happening at Criteo Labs.