Data and AI
Bol.com is a webshop in the Netherlands and offers general merchandising products in categories such as music, film, electronics, toys, jewelry, watches, baby products, gardening, and DIY. As of 2020, the store served 11 million active customers in the Netherlands and Belgium.
Data for Decision-Making First
Founded in 1999 as a pure internet company, bol.com doesn’t have the baggage of traditional organizations in becoming data-driven. The company’s advanced analyst Melissa Perotti explained, “Decision-making at bol.com is based on data first. The data doesn’t confirm someone’s hunch; it tells us where to go.”
As the largest e-commerce platform in the Netherlands and Belgium, bol.com has vast amounts of data for decision-making at its fingertips. “We use data science in many places—our search engine and recommender systems, forecasting of sales and customer service, and our chatbot. We have a mature data landscape and are proud of what we have achieved so far—and we’re ready to take it all to the next level.”
Over the past two years, bol.com’s data science capability quickly matured and now the shopping platform’s next challenge is to extend this beyond its data engineers and data scientists.
Keep Learning to Stay at the Forefront of Data Science
Being a frontrunner in the Dutch data science landscape is a great advantage, but it comes with its own set of challenges. Staying at the front, especially in a field that evolves as quickly as the data and AI space means never resting on your laurels.
Perotti elaborates: “Attracting great data scientists to work at bol.com means having a strong development program. We started ours about two years ago. We also founded a data science community within bol.com around the same time, to improve cooperation and knowledge sharing among data scientists, not just internal capabilities.”
For the data engineering and data science training program, bol.com looked for a partner to support their journey. The two main criteria for its program was applicability and being up-to-date with the newest technologies and models. After comparing different partners, bol.com chose Xebia. “Xebia understood how we saw our data science capability, and where we wanted to go,” said Perotti. “It was also really flexible in setting up the programs, so the courses were perfectly tailored to our situation, and the material was very applicable in our work.”
When bol.com kicked off the training program, it tested to see what worked best. “It was a challenge to ensure the applicability of a course where employees with different backgrounds come together. That learning process is still ongoing and we feel we’re getting better at it with every step we take,” explained Perotti.
A Wide Host of Courses
The in-company training program offered a wide range of courses to data professionals at bol.com. Courses included programs for data science starters like Python Essentials, but also more advanced courses such as Deep Learning.
Most of the courses are part of the Xebia training curriculum offers and were adapted to fit bol.com’s needs. For example, the A/B Testing and Experiments course was split into technical and non-technical training. The technical training focused on data scientists and ways they could apply experimenting in their day-to-day work. The non-technical training was aimed at managers, to give them an insight into the possibilities.
To provide training to the right employees, bol.com developed an “a la carte menu” displaying the different types of courses. This ensures that participants sign up for a training that fits their needs.
Extending the Training Program
After providing several courses for bol.com in 2018 and 2019, the program has been extended into 2020 with several additional courses. In 2020, Xebia will also offer the Data Science for Product Owners and Reinforcement Learning training, among others.
Gert-Jan Steltenpool, the sales director of Xebia Academy, has been heavily involved in setting up and supporting the program.
“What is great about the program with bol.com is that we really work together on the execution. After every training, we conduct an extensive evaluation to see if there are things we can improve from either side.”
Koen van Riemsdijk, Skills Accelerator, Xebia
Own Experience in Deep Learning
Participants at bol.com value the training courses, with an average score of 8.0. Perotti also attended several courses, with the Deep Learning course standing out for her.
“The training was a great combination of theory and practice. We worked with text models because we use these a lot at bol.com—for our chatbot, search engine, product descriptions, you name it. What I learned in the training was very applicable in my work.”
Melissa Perotti, Advanced Analyst, bol.com