Data strategy: purpose first, then the plan

This class was hard for me at first: during the first part I wasn't clear on where we were heading. In the end everything clicked, and it left me with an idea that works as well for a single project as for a whole team: before talking about data, you have to know what it's for.

Starting with the destination

Rafael opened the class with a quote from Seneca that I jotted down:

There is no favourable wind for those who don't know where they're going.

Every project, whether for a team or a whole company, starts with an objective. A strategy, he explained, is a purpose plus a plan. The purpose defines what value we want to create and how we'll measure it; the plan turns it into actions. The full path goes through seven stops, and this class focused on the two in the middle: diagnosing where we are and designing how to get there.

Data-driven organizations

A data-driven organization puts data at the centre of its decisions. What I liked most is that the starting point isn't technology but the question: they know how to ask the right business questions so they can answer them with data. These are the traits that come up most, in my own words:

• Most of their value comes from using data, and they know how to measure it.
• They ask business questions that data can answer.
• Their data is ready to use, not trapped in silos.
• They nurture the data talent they already have inside.
• They offer customers a different experience: personalized products, prices and communication, and the ability to anticipate.
• They use agentic AI as an accelerator.

Where are we? Control, optimization and transformation

The part that made me think most was a diagram for placing a data strategy according to its analytical maturity and its ambition. This is my drawing of the idea:

The three zones aren't better or worse, they're stages, and each asks for different things. What was uncomfortable was applying it to myself: I thought of projects of mine that I believed were quite advanced and that may not yet have made the leap they need to move from one zone to the next.

An example: applying for a car loan

To bring it down to earth, Rafael used the journey of a person applying for a loan to buy a car. Since I work in consumer finance, I recognised it right away. There are six moments and each uses different data: setting the strategic objectives, launching a personal loans campaign, receiving the application, deciding risk and price, disbursing the loan and, finally, reporting to the supervisor and measuring results.

The takeaway is that having good data at each step isn't enough. You need a data model that connects all the steps and that every team understands in the same way. That's where the whole class was heading.

From diagnosis to plan: two tools

To move from theory to practice, Rafael introduced two tools from the CDO Club Spain&Latam, created by data leaders from different industries:

• dataMat, the diagnosis. A survey of 105 questions, five for each of the 21 areas of the data world. Each company answers it honestly and gets a score in each area, to see where it stands.
• dataToolkit, the card game. One card for each of those 21 areas, which he gave us in class. On each one you place yourself at one of three levels: one point if you're in a bad place, two if you're so-so, and three if you're doing well.

An example with the data dictionary and glossary:

I liked thinking of this as a roadmap: the diagnosis tells you where you're failing and the plan orders the small changes to improve step by step.

What I take to my work

Review the data model of an area. Data models feed analyses, dashboards, predictive models and risk models, and the data usually arrives from technology to be turned into business intelligence. A good question is whether an area's model would meet those 21 areas. It's hard to extrapolate to a whole company, but starting small, one well-built model for one area, matters a lot: it helps tables make sense together and lets AI relate them and build coherent models.

Bring data closer to everyone. Rafael insisted that, in a data-driven company, even people who don't use data every day need to understand what is happening in the organization. And it has to be explained in their language: you don't talk about data the same way in every team. Building that data culture, at a level each person can handle, is an essential part of the job.

A project I'd like to do. Bring data closer to the people I work with, with three ideas:

• A data model of their own for each group, with their needs and a vocabulary we all understand.
• A simple, individual view that each person can check to understand their day-to-day and that also keeps them motivated.
• Listen before building: talk to people with different profiles, from those who perform best to those who need more support, to understand what they need and what motivates them.

Ecosystem, the keyword

If I take one word from the class, it's ecosystem. The leap to transformation comes when data models relate to each other in the best possible way and create an ecosystem where each area's data feeds the others.

What I take away

Before choosing technology, you have to be clear about what it's for. Then, diagnose honestly where we are and build a small, repeatable plan. And remember that data only creates value if the person using it understands it.

Thanks to Rafael Fernández Campos for the class this article is based on.