“When data helps, acceptance comes naturally.”

Between quick wins and long-term data strategy

‍

What does it take to make data-driven decisions? In a remote interview, data consultant and project manager Simone and data engineer Felix explain when building a data platform is worthwhile and when a bi-only approach is sufficient.

‍

Many companies want to work in a more data-driven way and start with dashboards. Is this the right approach from your point of view?

‍Simone: Yes, definitely. Dashboarding is often the best way to start because it makes data visible and tangible. When employees see how easily they can access information and how decisions can be backed up with real figures, acceptance increases enormously. As a result, data becomes part of everyday working life and that is the basis for every data culture.

‍

“As a result, data becomes part of everyday working life. ”

‍

Felix: I feel the same way. Dashboards quickly add value and show how powerful data actually is.

‍

What are the basic requirements for reporting to work reliably at all?

‍Simone: The biggest is clearly the data quality. If the data is incorrect, even the most beautiful dashboard cannot deliver reliable results. How do you say: Shit in, Shit out. In addition, a basic understanding of the data is required, otherwise errors will not be detected at all. And: Reporting is a process. It is normal for there to be disagreements at first. It is important to fix them instead of fleeing back to Excel.

‍

Felix: And then there is the organization. Companies must create the right roles and responsibilities, i.e. clearly define who is responsible for data quality, architecture and business logic. Otherwise, there will be gaps or duplication of work, and in the end no one trusts the figures. But dashboards are just the start. As soon as more data sources are added or more complex analyses are required, companies with a pure BI approach reach their limits.

‍

Many companies build dashboards before they even have a data platform. Can that still work?

‍Simone: For starters, yes. If you want to show something quickly or deliver initial results, this makes sense, for example to convince management. But it will be difficult in the long term. The data is then often scattered, there are no uniform definitions and performance losses. At the latest when data from several systems is to be combined, a solid architecture is needed in the background.

‍

Felix: Without a data platform, it becomes cumbersome at some point because every dashboard develops its own logic. A central data platform ensures consistency, reusability, and scalability. It is virtually the foundation on which companies can build sustainably. Regardless of whether for reporting, advanced analytics and later AI.

‍

“Without a data platform, it will be cumbersome at some point. ”

‍

Why is the data platform becoming so important when it comes to AI?

‍Felix: Because AI is only as good as the data it's based on. A modern data platform ensures clean, consistent, and up-to-date data. This is the basic requirement for every machine learning model. In addition, platforms such as Databricks, Fabric or Snowflake already offer integrated AI functions that can be used to train, monitor and roll out models. This lowers the entry hurdle enormously.

‍

Simone: And at the same time, a good data platform also helps analysts because it ensures that everyone works with the same, verified data. This strengthens trust both in dashboards and in later AI applications.

‍

‍

Which platforms are currently particularly relevant and what should companies pay attention to when choosing?

‍Felix: At the moment, Databricks, Microsoft Fabric, and Snowflake are the most widely used. The decision depends heavily on the team's skillset: If the team works a lot with Python, Databricks or Fabric are usually the better choice, while Snowflake is more suitable for teams with SQL experience. The use case also plays a role. Fabric and Snowflake are strong for classic BI scenarios, and Databricks develops its strengths in AI and machine learning. And, of course, the budget plays a role. The cost structures differ significantly.

‍

Simone: It's worth starting with a clearly defined use case and testing the platform in practice. In this way, companies quickly recognize what really fits their requirements before making a long-term commitment.

‍

How much effort does it take to build such a platform?

‍Felix: Of course, this depends on the scope, i.e. on the number of data sources, the complexity of the transformations and the desired level of automation. But companies don't have to build everything at once. Many start with an MVP, i.e. a minimum viable product that is ready in a few weeks and delivers real benefits. After that, they can gradually expand.

‍

Simone: That's right. It is crucial that visible results are achieved early on. This creates trust and motivation. When data makes everyday work noticeably easier and processes become easier thanks to the platform, acceptance grows all by itself.

‍

In your opinion, what is the decisive success factor for data-driven work?

‍Simone: For me, it's the combination of quality and culture: good data, but also people who want to understand, question and use it.

‍

“Data quality is and remains the basis. ”

‍

Felix: And an architecture that supports the whole thing: scalable, transparent and flexible enough to evolve. Data strategy, platform and people must work together. Only then will data work be sustainable.

‍