THE BI WORLD🌏IS FLAT🥮

img

Full disclosure. I believe the world 🌏 is flat 🥮.

More precisely, I believe the business intelligence world is flat.

Mostly flat in any case.

I believe the world can be flat-approximated to respond to most corporate reporting questions.

That this idea is actually correct is very important for me.

You see, this tool only eats one flat tidy file per run.

The situation is actually serious.

I spent years developing a tool essentially because I believe you can answer most corporate reporting questions with a measly flat tidy file.

If this is not true, my app is worthless.

Anyway, here goes the explanation.

I started doing reporting and analytics way back then in Excel as a management consultant.

The client would give a few Excel files with the company’s data to analyze.

I would invariably transform each file into a flat “tidy” (forgot to mention — tidy just means one row per observation, one column per dimension) dataset and then I would start number crunching and analyzing it. I would do some vlookups here and there, but, once done with vlookup-ing, the object of my analysis was always 1️⃣ lonely flat tidy file.

Same when I started playing with Tableau. Systematically one dashboard, 1️⃣ flat tidy file.

I refined this further when I started working with SQL and PowerBI for a large customer. I would do as much as possible with SQL, join a ton of tables directly in the ERP database, and make sure to end up with 1️⃣ — you guessed it — flat tidy file.

Very occasionally, I would end up with a star schema in my PowerBI. But it was the classic “fact file” star + “dim files” planets thing that can be reduced to 1️⃣ non-normalized flat tidy file.

OK, I know, corporate reporting cannot all be mapped with a flat file.

For example, if you want to understand sales in connection to the sales funnel, you will have, for each sale observation, one or more steps in the funnel.

You can’t use a single flat file, you need to connect two fact datasets with a join to do your analyses.

There are many such use cases.

That is why those two Italians invented data modelling in PowerBi. They wrote a book. I read it a few times. Repeatedly tried to understand it. Always failed miserably. I ended up with buggy implementations. That had to stop.

This is the story of why I decided that my reporting world is flat.

If I see a curved horizon, I will ignore it and keep my distance.