Do you need a data platform?
4 years ago I was a speaker for an event called "Power BI for municipalities". The first speaker showed some of the reports they had been making at their organization. The audience was very involved and enthusiastic. They simply loved it. The main feeling going through everyone's mind was "I want to start with Power BI in my organization!".
I was up next. And my message was basically "Power BI & self service is great, but you probably need a data platform".
Talk about buzz kill. You felt the energy change as all the people involved suddenly began to rethink their first conclusion "so....it's complicated after all?". They did not want to know that! Reading this, you may wonder: "why bring up the subject again Jesse?".
The reason for that is that people get easily confused with Power BI. It's originally a self service BI tool. Aimed at business people to create reports for themselves and others. Something like Excel, but solely focused on reporting. On the other hand, it can also be the frontend your enterprise reporting platform, delivering data from the data warehouse and used for critical decision making.
See the potential confusion? Reports coming from self service and enterprise reports end up in the same platform: www.powerbi.com. It can be difficult for the users to understand this. Lately the confusion can be even greater; with Power BI premium, business users can create large data models for thousands of users. With Power BI data flows, business users can create large ETL solutions. If the business can deploy self service solutions to thousands of users, what is the difference between enterprise and self service anyway?
The difference between self service BI and enterprise reporting needs to be clearly defined for an organization that wants to use Power BI. For me, self service BI is people creating reports for themselves and a select few others, that is not classified as an official report to base company critical decisions on. Enterprise reports are the official classified reports that are cleansed and are suited for decision making. Typically, self service is created by business users and enterprise reporting is created by IT. Since the skills start to overlap more and more, I don't think that is the most important thing in the definition anymore.
So back to the initial question, do you need a data platform? Let's first define what it is: A data platform is the combination of solutions (datawarehouses, datamarts, data lakes, solutions, data cleansing, data flows and cubes to name a few) to deliver historical reports to the business. The key here is that it is organized. It may actually be all virtual, it does not have to be persisted.
For business critical reporting you may very well need a data platform. You may need to keep a record of the raw data before it is cleansed for auditing purposes for example. Or you need to have a single consistent view across the organization. To organize that all with Power BI is almost impossible.
With Power BI the report is tightly coupled to the data model. Both a data model and a report are created in a single pbix file. Now it is possible to create other reports on top of the data model located in another report, but this is not always so easy and often, the very nature of Power BI makes it easy to have 60 data models for 60 reports. Meaning that if you have multiple source systems, you very likely may end up with 60 ways of integrating them. Having a data platform where that is already done makes a lot of sense.
- Do you have a lot of unstructured data? You may need a data lake.
- Do you have a lot of source systems that need to be integrated consistently? You may need a data warehouse.
- Do you need to apply machine learning at scale? You may want to look into Azure Databricks.
- Do you need to track changes to the solution? You may need source control.
Power BI is great, but it is not yet the single tool that can do everything.
Have anything to add? Leave it below in the comments 🙂