Building a Data Science team in marketing (a field I had to learn from scratch)
Lessons from stepping outside my expertise to create a high-impact marketing data science team.
This week, I have been reflecting about an unexpected opportunity I was given in 2022: building a data science team from the ground up.
I had been working at Skyscanner for a few years (mostly in data engineering and recommender systems). But then, I was asked to build a marketing data science team from scratch. The issues?
First, there was no existing data science team in marketing (nor had there been in the past).
Second, I myself had no prior experience in marketing.
And third, there was no clear roadmap on where to start (yes, stakeholders wanted Data Scientists, but no one knew how we could contribute).
As you can see, this was not just another technical project. It was about navigating new stakeholders, understanding the marketing landscape, proving value... And proving it fast.
Over the next couple of years, I built a high-impact marketing data science team from 0 to 6 data scientists.
In this article, I wanted to share my learnings and walk you through:
How to structure and scale a marketing data science team.
How to prove value quickly (without getting stuck in analysis paralysis).
The key mistakes to avoid when integrating data science into marketing.
If you are a data leader, marketing exec, or just curious about how data science drives marketing decisions, this will be useful.
Let’s get started!
Historical context
Why did Skyscanner invest in a marketing data science team?
At the time, Skyscanner’s data science teams were deeply embedded in four core areas:
Flight recommendations and ranking – optimising our core product.
Search engine optimisation – improving user search experiences.
Experimentation – running A/B tests at scale.
Internal ad platform growth – improving monetisation strategies.
We were a 20-person data science team proving our value across these areas. But, there was a big area where data science wasn’t part of: marketing had no dedicated data science function.
The company was spending millions in paid ads, and someone thought that if spend was to be optimised, you needed machine learning and data scientists.
Why marketing?
Marketing is one of the most powerful levers for growth. If you already have a sticky product that users love. “All you have left to do”, is bring more travellers through the door… Sounds simple, right?
However, you can’t just simply pour more money into this. The challenge is to spend money efficiently. If you go wild with money spending and to look for more customers, the customer mix changes (you are tapping into customers who might not know your brand, those who don’t convert well or even spam traffic).
And so, I was given a clear (but vague) mission:
"We need a marketing data science team to drive sustainable growth."
And just like that, the journey began…
Phase 1. Identifying the biggest problem to solve
Before diving into hiring or execution, I needed to answer one critical question:
💡 Where could Data Science drive the biggest impact in marketing?
Without a clear answer, the team risked becoming a nice-to-have function instead of a must-have growth driver.
In phase 1, my goals were:
To understand the entire marketing ecosystem. Identify where Data Science could add the most value.
To create a decision framework. Assess each team's gaps, feasibility, and potential ROI of data science efforts.
To build trust with key stakeholders. Without buy-in, even the best projects fail.
To define a clear mission statement. So the marketing team knew exactly how Data Science would help over the next two years.
This was not just about picking a project.
It was about strategic positioning.
Mapping the marketing ecosystem: Where does Data Science fit?
Before deciding where to focus, I needed a clear picture of how Skyscanner’s marketing function operated. Marketing at Skyscanner was big, more than 100+ people across four core pillars:
🔍 SEO – Driving organic traffic growth.
💰 Paid marketing – Managing ad spend across multiple channels.
🤝 Affiliate growth – Expanding partnerships with external sites.
🔄 Lifecycle marketing – Improving retention and customer re-engagement.
Each of these areas had dozens of projects running simultaneously. But not all of them needed Data Science. I needed to map out the landscape to figure out:
Where Data Science could drive the highest impact.
Which teams were far from their targets and would benefit most from our help.
What was realistically feasible given our limited resources.
This ecosystem audit became the foundation for our decision-making.
Focusing on impact: The T-shirt sizing framework
I knew one thing for sure: if I wanted to build a high-impact data science team, I needed focus. This was not about spreading thin across all marketing areas. It was about creating a centre of excellence in one domain first, proving value, and then expanding.
Given the many existing projects, I knew I couldn’t serve them all. If I wanted to scale this new team to 6 people, I wanted the 6 of them living and breathing under the same pillared umbrella. The time to expand to other areas in marketing would come, but adding value and making a reputation of what we could do was the most important thing for me at the beginning.
This meant saying no—and that was not always easy.
Some teams were frustrated when we decided not to allocate Data Science resources to them. To keep things objective, I introduced a simple but effective decision framework: the T-shirt sizing model. This framework helped us prioritise marketing areas based on:
Help required – How much does the team depend on Data Science?
Potential opportunity – How much impact can we create?
DS efforts – How complex is the problem to solve?
Note that the “Potential opportunity” and “DS efforts” column include figures such as 1m or 2 quarters. Take these as “orders of magnitude” references. My goal was to differentiate a S or XL t-shirt size easily.
By structuring our decisions with clear criteria, we were able to avoid politics, focus on impact, and set the right expectations across teams.
Why Paid Marketing? I chose the highest-impact problem
After applying the T-shirt sizing framework, one marketing pillar stood out: Paid Marketing.
Here is why it was the highest-priority area for Data Science:
🔥 Help required (XL): Skyscanner had nailed SEO (our organic traffic was strong). But paid marketing was an untapped opportunity: “How can we automate bidding to grow paid traffic in a sustainable way”.
💰 Potential opportunity (XL): Skyscanner was spending millions annually on paid ads. If Data Science could optimise this spend, even a small efficiency gain could mean millions in additional revenue (or saved costs)
⚙️ DS efforts (XL): This was not an easy problem. At the time, we had zero existing Data Science tooling for marketing solutions. We were new to this space, and any project would require deep integrations with marketing systems, building automation from scratch, significant trial and error to get results.
But despite the high effort, the return on investment was too big to ignore.
Why we didn’t focus on SEO or Lifecycle Marketing
Here is why we didn’t allocate Data Science resources to SEO or Lifecycle Marketing (at least initially…):
SEO: A high-impact but well-resourced area.
They were already on target with their goals.
Technical expertise was strong, and they had the right tools to execute.
Even though a 1% improvement in SEO traffic could generate millions in revenue, the biggest bottlenecks were already being addressed.
Verdict: SEO was important, but it didn’t have an urgent Data Science gap.
Lifecycle Marketing: Already supported by other teams
Lifecycle marketing was another tempting area, because they had already worked with Data Science. This means that my transition would have probably been pretty smooth. Still, their needs could be handled by existing data science support.
Verdict: The opportunity in Lifecycle Marketing was smaller compared to Paid Marketing’s tens of millions in ad spend.
While both SEO and Lifecycle Marketing were important, the biggest gap and highest ROI were in Paid Marketing. That is where we placed our first bets.
Building relationships with Paid Marketing: Winning stakeholder buy-in
Having chosen paid marketing as our data science focus, it was really important for me to get a solid working relationship with the relevant stakeholders.
Here is how I built strong working relationships with key Paid Marketing stakeholders:
Regular 1-on-1s with senior decision-makers To drive impact, I needed to connect with the people making the hard trade-offs. Every 2-3 weeks, I met with Senior Directors across:
Marketing (to understand growth objectives).
Product (to align on data-driven strategies).
Engineering (to ensure feasibility and execution).
These meetings helped me understand priorities while building trust.
Weekly project check-ins with cross-functional teams. Beyond senior leaders, I needed to keep execution teams aligned. I set up weekly calls with Product, Engineering and Marketing. Each week, we reviewed:
A 1-quarter roadmap – Clear deliverables and progress tracking.
Key blockers & dependencies – Identifying risks early.
What help we needed – Ensuring teams were properly resourced.
These rituals kept everyone on the same page—from leadership to execution teams.
Defining our mission statement: What was our north star?
After aligning with stakeholders and setting priorities, we needed a clear mission statement for the new data science squad. Our mission was simple, but powerful:
Optimise our most important bidding solutions to grow Skyscanner’s audience in a sustainable way.
This mission did two things:
✅ Kept us focused on high-impact problems (bidding automation, not scattered initiatives).
✅ Aligned with Skyscanner’s long-term marketing strategy (growth + sustainability).
By defining this early, we set expectations and ensured everyone knew what Data Science was here to do.
Phase 2. Proving value - winning our first big bet
We had a mission. We had stakeholder buy-in. But we still had one critical hurdle to clear: could Data Science actually deliver impact in marketing?
To prove this, we needed a high-visibility, high-value project. We chose Hotel Price Ads.
With this in my mind, my goals for phase 2 were:
Hire the first two data scientists.
Set expectations with stakeholders: The first quarter would be about building foundations, not immediate wins.
Build tooling & automation: Could Data Science be self-sufficient in production deployments?
Prove value with a strong proof of concept (PoC). If we demonstrated success, we could unlock more headcount and projects.
This phase was about laying the groundwork, without necessarily rushing for instant results.
The chosen project: Automating bidding & supply selection
Paid marketing teams were manually managing bids and supply selection. This was:
Slow, as adjustments were reactive, not proactive.
Non-scalable, as execution required on a paid marketeer to be changing bids through an UI.
Blocker, as Tripadvisor’s bidding system required massive CSVs as inputs. The best shot at getting those CSVs prepared was doing it programatically.
Data Science was tasked with growing our acquisition of hotel users through Google and Tripadvisor. We developed:
Bidding algorithm: Decides how much to bid for a specific hotel + itinerary combination.
Supply selection algorithm: Determines which hotels & itineraries are worth bidding on.
Scalability was non-negotiable. If we only proved value in two markets with manual adjustments, it would not have been a true Data Science solution.
These 3 had to be fully programmatic.
Our goal was clear: Machine-led bidding & supply selection that could scale globally.
Did we actually prove value? Check the results from our bidding algorithm in Hotel Price Ads
Talk is cheap… did our Data Science models actually move the needle?
In the following screenshot, you can see a snapshot of how our work looked like in terms of transactional growth coming from Tripadvisor.
With this, I want to show you how the project evolved, and why it was important to keep constant communication with stakeholders.
Phase 1: The learning curve (messy but expected)
ROAS and volumes were all over the place.
We were still integrating with marketing systems and testing early ML models.
This phase was about finding out what worked (and what did not).
Stakeholders knew at each point in time, what sort of ideas were we tested and the goals for each test.
Phase 2: Hitting our stride
We gained clarity on key optimisations for the next quarter.
Aligned ROAS with targets while significantly increasing conversions.
Proved that Data Science-driven bidding could outperform manual approaches.
Conclusion?
We had proof that Data Science could drive real marketing impact.
This success opened the door to more headcount, more projects, and deeper integration into paid marketing strategy.
Phase 3. 3x team growth to own 3 projects.
We had proven our value in Hotel Price Ads. Now, it was time to scale.
In phase 3, my main goals were:
Grow the team from 2 Data Scientists to 6
Scale our existing hotel bidding project with best MLOps practices.
Open 2 new projects in the bidding space.
Having proven that Data Science could add value to the paid marketing space, and being much more knowledgeable in the area, we were ready to open new projects and scale the team to 6 Data Scientists.
My goal was to ensure that no data scientist worked on their own, so we paired up to work on 3 different projects:
Optimising bids for hotel price ads (existing project, but keep improving our bidding algorithms whilst opening new partnerships).
Optimising bids for SEM. This was super big, and conceptually, similar to our first project. This allowed us to hit the ground running on the tooling we had. But, we did approach SEM problems differently to hotel price ads, so it took some time to show the same sort of growth results.
Optimising paid APP installs. This is a massive project, because we are not only dealing here with a bid optimisation problem, but with a privacy problem. iOS and Android were tackling the privacy world, so we could not pin down the relationship between our internal transaction and the specific upstream click we were buying. This is still an ongoing project and most probably one which we will need more DS brain power behind.
By the end of Phase 3, we had evolved from a small experiment into a fully embedded Data Science team in Paid Marketing.
Final takeaways: How we built a high-impact marketing Data Science team
Building a marketing data science team from scratch was never about fancy ML models or quick wins. It was about:
✅ Finding a problem space we could own. Instead of spreading thin, we focused on bidding automation in paid marketing.
✅ Proving value early. Hotel Price Ads became our first major success, unlocking more headcount and trust.
✅ Structuring the team for scalability. We made sure every project had a clear strategy, ownership, and long-term vision.
✅ Building strong relationships with stakeholders. Without senior leadership buy-in, none of this would have been possible.
The biggest lesson?
If you are building a data science team, do not chase impact everywhere
→ Own one area.
→ Prove value
→ And scale from there.
Your turn
Have you built a Data Science team? What challenges did you face? Let me know in the comments!
If you also found this useful, share it with someone who is building a Data Science team!
Further reading
If you are interested in more content, here is an article capturing all my written blog posts accessible as a paid subscriber.
