We are often asked what kind of work Product Data Scientists do at Criteo. Julie and Solene have agreed to answer some questions, giving us insights on their day-to-day work as Product Data Scientists at Criteo.
Fresh out of CentraleSupélec and Columbia University, we both decided to join Criteo, as a first job. Criteo seemed to be an amazing place to continue learning while working in a global, diverse, dynamic and fun environment. Spoiler alert, after 1.5 years at Criteo, we can safely say that it did not only seem but it really is. Working as Product Data Scientist allowed us to have access to a wide range of products and teams across Criteo, combining technical challenges to business impact.
What does the Product Analytics & Data Science (PAX) team work on?
The Product Analytics & Data Science team works on every aspect of Criteo core-business and product portfolio. We have the chance to meet with many different teams across Criteo. We work closely with R&D, Product Managers and other Analytics teams like Local Analysts (facing marketers), Global Analysts (developing internal tools) and Global Supply Analysts (facing media owners).
Currently, PAX is divided in four teams focusing on specific Criteo products. Each team has its own expertise and works closely with the others to get the best possible outcomes.
The goal of PAX is to participate in product creation and improvements. We participate in various projects like the launching a new product andexploring strategic game-changers or keeping our existing products competitive and groundbreaking to vault Criteo ahead in a fast-evolving media landscape. In all situations, we use data to make decisions and always keep a client (retailers, brands, agencies, publishers...) and user centric approach in our work. We bring data and business expertise to feature teams (often a combination of Engineering, Product Management and Analytical ones) and a unique understanding of complex Criteo machinery to design and build products with difference.
What does your usual day look like?
In a typical day, we work on long-term projects defined every quarter with our manager. We take complex and critical challenges, transform them into complete analytical projects, to give actionable insights and drive the product development.
Julie: for example, one of my projects is finding and testing replacements to third party cookies. I design offline tests and AB tests to evaluate the performance of new bidding mechanisms. Offline tests give preliminary results and estimations to refine the scope of the AB Test which is usually jointly launched with R&D. I work closely with Product Managers on this topic. They bring overall strategy for a product development, and I bring data arguments and quantitative goals. It is great to have an impact and take part in product decisions.
Solene: One of the projects I really enjoyed was Ad Safety where I worked very closely with R&D. The goal of the project is to ensure that images displayed in Criteo Ads are compliant with local regulations. To achieve this, R&D has developed a model to identify images that are unsafe (guns, nudity, drugs... you name it). We rolled out the model, and computed model performance and uplift compared to a benchmark. I conducted an AB test during the roll-out, identified patterns in misclassification to help improve the model and size the opportunity of combining the image model with a text model. It was interesting to work with developers and have hands on deep learning model at scale.
The Product team is the cornerstone of Criteo. We collaborate with a lot of other teams. As a result, we spend between 1 hour and 3 hours in meetings every day. Usually, a 25 minute daily catch up with our team to update each other on our work advances.
We meet with R&D teams working on the same scope to synchronize and clarify the next steps in our projects. It is great to be involved in technical discussions with developers and show them data insights on the products they work on. It is also a great place to influence roadmap with product needs and objectives.
We also have weekly presentations dedicated to knowledge sharing between PAX teams. These meetings are a great opportunity to practice communication skills and have a broader overview of what other PAX members work on.
What kind of tools do you use?
Criteo is a tech company, which means the amount of data is not only huge (to give you an idea, Criteo ingest 700Tb a day) but also well organized and accessible. As data enthusiasts, we all agree this is the dream for Analytics and Data Science. The source of all our data are the raw logs generated by Criteo applications. We usually access it through SQL queries using Presto or Hive engines and via Interface, terminal or python, depending on the use case. There are several R&D teams ensuring the data is transformed and aggregated for our usage. We are also free to create our own aggregated tables using custom jobs.
For reporting needs, we like using Tableau. The chart & graph portfolio is huge and very suitable to our use cases. Plus, we can publish our dashboards online and make them available from any Criteo laptop. This way, commercial and technical teams have access to relevant data points updated daily.
We use Python a lot for ad-hoc analysis, data exploration and Data Science projects. There is the amazing CriteoPy Python Package - developed by Criteos for Criteos - to which everyone can contribute. Some of us also use Scala occasionally.
One more cool thing, we have a bunch of custom tools developed internally at Criteo. For example, if you made a change on an aggregated table you own, you can very easily let users know what changes were made using a tool sending Slack messages to impacted people. It makes our work so much easier and fun.
What do you enjoy the most about working at Criteo?
Both of us have the same answer to this question. The people. There is an amazing diversity at Criteo, the company culture is real: open, together and impactful. In all the interactions we have at Criteo, no matter the team, we always meet helpful, kind and extraordinarily talented people.
The company supports employees to be themselves and takes exceptionally good care of us. As of today (June 2021) we are working from home. Criteo has been super reactive to Covid19 and we were able to work remotely through the entire pandemic with support of the company. We managed to maintain social link with virtual drinks and games. The company and co-workers will help grow a lot, with trainings, bootcamp and peer review cycles. there is the opportunity to do a “voyager”, i.e. work on a project with another team, if you want to discover another part of Criteo and another job, for a few days, a few weeks or a few months.
Criteo is , investing on research and new project exploration. In PAX, we are working in an agile fast moving environment but with the means, variety of talents and scale of a large company. Criteo is an exciting and very motivating company to work for.
The Paris office is also a great plus. We are located Rue Blanche between les Grands Boulevards and Pigalle, a very dynamic neighborhood. We usually have coffee as a team in the morning, at lunchtime we have an amazing cafeteria where we all go together. Not to mention the 360° rooftop, nap or game rooms and afterwork drinks.
From the inside
Discover how Sebastien's mentoring experience has helped him diversify his skillset.View More