Building a Strong Data Science Team Culture

[ leadership datascience lazada ] · 5 min read

I know, I know. I’m guilty of not posting over the past four months. Things have been super hectic at Lazada with Project Voyager (i.e., migrating to Alibaba’s tech stack) since last September and then preparing for our birthday campaign in end Apr. In fact, I’m writing this while on vacation =)

One of my first objectives after becoming Data Science Lead at Lazada—a year ago—was to build a strong team culture. Looking back, based on feedback from the team and leadership, this endeavor was largely a success and contributed to increased team productivity and engagement.

Why culture?

When I first joined the Lazada data team, we had 4-5 data engineers and data scientists combined. A year later, we grew to 16. After another year, we were 40-ish. During 1-on-1s with the team, some of the earlier team members raised concerns that our culture was being diluted as we scaled, and it “didn’t feel the same anymore”. Back then, different team members had different views of what our culture was.

In addition, during interviews, many candidates would ask about our culture—this was key in determining if Lazada Data Science was a good fit for them. Having a culture document available for sharing before interviews allowed candidates to learn more about us beforehand, and was more scalable (than answering questions at interviews).

Given the above, I made drafting a Lazada Data Science culture document a priority. It would indicate to current (and future) team members the values and behaviors that were encouraged and provide clarity on our working environment. In addition, we wanted to openly share with potential team members to give them a better sense if we would be a good fit before the interview.

How did we draft this document?

First, I researched the culture of companies I admired. This includes:

  • Netflix for quickly pivoting from mail-order DVDs to streaming
  • Valve for moving from games to platform
  • Google for managed chaotic innovation
  • Amazon for customer obsession and results

I tried to understand the intent behind each of their values, why they chose those values, and how it worked well for them in their context (e.g., market, stage of growth).

I also consulted internal engineering and data leaders on what they felt was essential culture for high performing teams. This was helpful as they shared deeply about the good and bad aspects of Lazada’s culture (e.g., moving and failing fast—good, not taking ownership—bad). In addition, I sought the opinions of my informal mentors—data and technical leaders in other companies—on what worked well for them. The research findings and consultation outcomes were largely similar.

After creating a draft, focus groups were held with the team and leadership to share informally and seek feedback. Everyone largely agreed with the draft and there was not as much debate as I expected.

Key points in Lazada’s Data Science team culture

Here’s a summary of our culture document. The full document can be found at the end of the post.

Our culture is built around five values:

  • Ownership: You behave like a founder, in buyers’, sellers’ and Lazada’s best interest, even against popular opinion—nothing is “not your job”.
  • Collaboration: You collaborate with other functions and the team to improve overall outcomes (e.g., make time for others, respecting diversity).
  • Communication: You are transparent with progress, results, and learnings (i.e., mistakes) and provide timely feedback so we can all improve.
  • Innovation: You generate ideas, challenge assumptions, and apply technology and data science advances to create practical and measurable value.
  • Impact: You ship. Also, you set ambitious targets, deliver consistently strong outcomes, measure them obsessively, and are a team lynchpin.

The following aspects are also an important part of our culture:

  • There are no mistakes, only opportunities to learn. In tech, our biggest threat over time is lack of innovation—everyone has the freedom and space to fail. “Failure is an option here. If things are not failing, you are not innovating enough”—Elon Musk
  • Everyone has the responsibility to do the right thing. Leadership provides the resources and context so everyone can perform good, independent thinking. The aim is to be highly aligned and loosely coupled, so we can all be fast and flexible.
  • Hiring well is the most important contribution to our success—we only hire people who are, or have the potential to be, better than us. Questions to ask ourselves: “Would I want this person to be my boss?”, “Would I learn a significant amount from this person?”. Beyond merely technical skills, we value initiative and teamwork.

Final thoughts on Culture

Some people feel that culture is hindsight bias—only after companies are successful is culture cited as a reason for success. There may be some truth to this, and there are companies that were/are very successful that do not have a good culture (e.g., Enron, Uber, Goldman Sachs).

On the flip side, others feel that culture is the most important factor distinguishing the best from the merely good. Often cited is the famous quip by Peter Drucker—“Culture eats strategy for breakfast”.

Having been in technical roles across government, giants, and startups, my view is that a good and strong culture is essential—it helps the team do their best work and be successful, grow in a safe environment, and attract more talent, which contributes to further success and the talent flywheel. Nonetheless, it is not the only necessary ingredient. Other aspects such as strategy, product, customer experience, are also key.

Thanks for making it to the end of this post. Here’s the full culture deck.

If you found this useful, please cite this write-up as:

Yan, Ziyou. (May 2018). Building a Strong Data Science Team Culture.


  title   = {Building a Strong Data Science Team Culture},
  author  = {Yan, Ziyou},
  journal = {},
  year    = {2018},
  month   = {May},
  url     = {}

Share on:

Browse related tags: [ leadership datascience lazada ]

Join 6,700+ readers getting updates on machine learning, RecSys, LLMs, and engineering.