“Trying to solve every problem at once is the best way to solve none of them well.”

When I led my first data team in 2010, I didn’t know this yet. I thought the job was to build the perfect system.

I was wrong.

And it nearly sank one of the most ambitious global data projects I’ve ever worked on.

The Trap Most Data Teams Fall Into

If you work in data you’ve probably been asked to “make it all work” at once.

Finance wants reports on profitability.
Marketing wants campaign attribution.
Ops wants warehouse stock data.
Everyone wants “the whole thing” right now.

But building a data infrastructure that solves all use cases from day one is a trap.

This is the story about how I learned that lesson the hard way and what you can do to avoid making the same mistake.

How I Nearly Broke a Global Data Project at 29

In 2010, I joined Rocket Internet. At the time, it was the largest venture builder in the world. Rocket had just launched Zalando, which became Europe’s leading online fashion platform.

Zalando’s growth wasn’t just fast, it was surgical. They knew how much return they would get for every dollar spent on marketing. That confidence in spend came from superior data systems.

Our plan? Replicate Zalando's data-driven growth in emerging markets across the globe through a project called Bigfoot.

I was 29, ambitious, and terrified. My job: build a global data infrastructure to power Zalando-style growth in new markets like India, Russia, Latin America, Africa and Australia.

And I made a big mistake.

The problem: Trying to build everything at once

We aimed to launch with a system that covered every major area of ecommerce:

• Acquisition marketing: Campaign performance, spend allocation, ROI modeling
• Retention marketing: Email engagement, churn prediction, customer segmentation
• Warehouse and stock: Reconciliation across multiple physical locations
• Purchasing: Stock replenishment and demand forecasting

At the same time, most source systems weren’t even stable yet. Many were still being built. The data was unreliable. The infrastructure was fragile. And yet, we tried to plug it all in at once.

What we opened was Pandora’s box.

The system didn’t deliver. Too many moving parts. Too much complexity. Too little focus.

I learned the hard way: chasing completeness early on kills momentum, frustrates stakeholders, and burns out teams.

The learning: Focus beats perfection

Looking back, there’s a clear answer to what we should have done:

Start with one stakeholder. Solve one problem extremely well. Then expand.

In our case, that would’ve been replicating Zalando’s acquisition marketing engine.

That engine was already proven to create real business impact. It was also highly measurable: perfect for a first use case.

Here’s the framework I use today, shaped by that experience:

The Minimum Viable Data Infrastructure (MVDI)

1 Understand all the needs​
Interview stakeholders. List all the use cases. Don’t skip this.

2 Identify the business priority​
​One priority. Not two. Not three. What’s the most important business goal right now?

3 Map the highest-leverage use case to that priority​
Often it’s the marketing spend engine. Sometimes it’s retention. Pick the one with a clear link to impact.

4 Build only what’s needed for that use case

• Limit data sources
• You might not need even need any dashboards
• A Google Sheet connection to one table might be enough
• Or feeding one piece of data into Google Ads
• Focus on correctness and usability

5 Make it work. Really work.​
It’s not called minimum viable because it’s sloppy. It’s minimum because it’s sharply scoped.

6 Then expand.​
Solve the next use case. Reuse infrastructure where possible. But always maintain focus.

A better definition of MVP (especially for data teams)

There’s a common misconception that Minimum Viable Product (MVP) means “bad version of a big thing.”

That’s wrong.

A good MVP does one thing very well. It’s viable because it solves a real problem with real quality, even if it only solves one.

If your MVP solves five problems at 20% quality, it's not viable. If it solves one problem at 80% quality, you're on the right track.

The bottom line

The perfect system is a myth. A dangerous one.

What your stakeholders need isn’t everything. They need something that works. Something that gives results. Something they can trust.

The way to get there is to:

• Ditch the “perfect system” mindset
• Focus on one clear business priority
• Build a data product that solves one high-impact problem well
• Deliver that, then move to the next

Done is better than perfect.

But more importantly: focused is better than done.

Your job is not to please everyone. Your job is to create business impact. Start small. Go deep. And grow from there.

Cheers,

Sebastian

💡 P.S.: Curious how to build minimum viable data products with YOUR data team and make your stakeholders and CEO love you? 😍

In my masterclass "Create massive business impact with your data team" I am sharing all my frameworks from 17+ years of building data infrastructure and crafting data strategies.

Check it out whenever you're ready!