I've built data organizations from zero five times. Best Buy. Target.
C.H. Robinson. Olo.
Shipwell. Every time, I made the same mistakes first. Then I stopped making them.
Here are 5 things most leaders get wrong when building a data team from scratch. 1. They hire technical skills first.
Your first hire shouldn't be your best coder. It should be someone who can translate between the business and the data. I've seen brilliant engineers build models nobody asked for.
At C.H. Robinson I built from zero to 45 people. The hires who survived year one weren't the strongest technically.
They were the ones who could walk into a carrier negotiation and understand what was really happening. 2. They build dashboards before building trust.
New data leaders love to show quick wins through dashboards. The problem is nobody trusts the numbers yet. At Best Buy we built beautiful reporting that executives ignored because they didn't trust the underlying data.
Spend your first 30 days fixing data quality and aligning definitions. Boring. But it's the foundation everything else breaks without.
3. They pitch technology when the C-suite wants outcomes. I've presented to boards over 20 times across my career.
Not once did a board member ask what tools we were using. They asked what revenue we were generating and what decisions we were improving. At Shipwell I secured a $3M budget increase by presenting ROI quarterly.
The slides had zero architecture diagrams. All outcomes. 4.
They centralize everything or decentralize everything. Both extremes fail. Full centralization creates bottlenecks.
Full decentralization creates chaos and no standards. At Olo during IPO prep I ran a hybrid model. Centralized data science standards and governance.
Distributed analytics ownership to product and engineering teams. That structure scaled under the most intense scrutiny a company faces. 5.
They skip governance until it's a crisis. At every company I've built a data org, I established governance from day one. Not because I'm cautious by nature.
Because I watched other companies scramble when a model failed, a bias audit surfaced problems, or a regulator came knocking. Governance isn't bureaucracy. It's what lets you move fast without breaking trust.
Here's the pattern underneath all five. Leaders build data teams like technology teams. The ones who succeed build them like business teams that happen to use technology.
That one sentence changed how I hire, how I prioritize, and how I present to leadership. It's the reason my teams have generated $2.5B+ in revenue impact. Build for the business.
The technology will follow. What would you add? What did you learn the hard way?