A few years back, I entered the office for my first day on a new job. The very first question I got was, “Do we keep the data warehouse?” It cost half-a-million euros to maintain, and there were millions of sunk cost because an isolated staff of analysts had worked for years at it. But nobody seemed to know what to actually do with it.

I decided to start talking to the analyst who was operating the system. I quickly noticed he was an extremely intelligent man. Nonetheless, our conversation did not go smoothly. He kept telling me what data was available. Look at that connection with the retail information. We can visualise things, he said.

But every time I asked him how we could put this data to use, he told me that was an inappropriate question. That deeply puzzled me. Inappropriate?

Having data is not enough. Knowing what to do with it is more important.
Having data is not enough. Knowing what to do with it is more important.

Then I went on to talk to the marketers. They said they had heard the data warehouse could do great things, with lots of connections. But they were not using the data and didn’t know what to do with it. Not invented here. How could this have happened?

After a week, I decided to get rid of the entire data warehouse since it was only costing us money.

Now I understand what was going on back then. And I see the problem with a lot of organisations.

There is small analytical group that understands the power of data. They receive a budget, lock themselves up in a room with other specialists, and come out a few months later, telling the business they have this cool thing that will improve their work.

But the business is too busy selling stuff, visiting clients, producing the product. They feel disconnected. And since the business people have trusted their gut for so many years, they will not change their inner compass for some wise guys saying they know better about the business.

If we take a step back, what is happening here? All too often, the analyst talks about the data architecture, the variables, the connections. There is a lot of talk about “what.”

Some analysts take it one step further. They come closer to the “why” of data. They talk about the possibilities, the insights. This is what analytical people derive energy from — understanding that one single truth: “Why is this happening?”

Once they know why, they are satisfied. Their bucket is filled to the rim. This is the reason so many analysts tell us what the organisation could do with data. Because most analytical people derive energy from the insight itself.

But that is not good enough.

Typically, organisations that are doing data right almost never talk about what you can learn from it. They talk about what data has done for the business.

The most effective data-driven organisations, therefore, turn things upside down.

First they look for intrinsically motivated business people with an analytical mind. Then they put them in charge of data teams, because it is not possible to outsource deep learning to other departments. Watch for analytical people among your project managers, marketers, and sales people. What are they doing after work? What did they study? Then take that giant leap and put an analytical business person in charge of the business data.

Centralised data factory versus decentralised data team

There are two ways to design a data-driven organisation:

1. Build a centralised data factory. This will bring great benefits. It will be very efficient because you can have specialised analysts. One for data architecture, one for database building, one for reporting, one for descriptive analysis, and one for predictive, which requires more statistical skills. Then there are specialists in visualisation. You can put the best of the best on each division of labour.

The problem with centralised data teams is that they tend to focus on the wrong things. Because the Achilles’ heel of any central data department is the communication with the business, these teams of highly intelligent people can be so disconnected from the real business issues that they produce fantasies — a predictive churn model for the sake of a predictive churn model with no rescue strategies attached.

Then the central factory produces fantasies very efficiently. If central data teams are a stubborn reality in a company, the best thing to do is to appoint a common sense person to interact with the business constantly.

But there is another way, which is followed by highly effective organisations ...

2. Put business people on data analysis and keep them extremely close to the business teams — preferably inside the business teams. If these data people are truly empowered and feel the freedom to follow their own course, they will ask the right questions and transform these questions into actionable insights.

It is only the last part that adds value. Many forget that insights are worthless if you do not turn them into better products and better experiences for your customers.

These decentralised data specialists can be viewed like one-man armies. “Data Rambos,” one telecom analyst calls them. They are not experts in every field of analysis, but they can sure handle the basics of every part of the field and will make sure they are very effective. Data Rambos tend to create control groups and check whether the assumption is proven in real life.

There are, however, some difficulties to overcome with this approach.

Sincere interest in IT

The first challenge is to get business people to learn about relational databases, reporting, visualisation, and predictive analytics.

In every organisation there is this area that is undefined. Is it the IT department’s or is it business? Since business is responsible for the end result, it is crucial that business people take the first step and learn the lingo of IT, the systems, and the people.

Even if there is ample documentation, which many times is not the case, asking IT to make drawings is usually a good start. It can also be helpful to start small, multi-skilled project teams on easy assignments.

Start defining. What is a unique customer in the database? Are there packages? Be friends with finance, but marry IT, a UK publisher once said. If IT and business build a relationship, the company will reap the benefits of this mutual understanding for a long time to come.