Leo Godin
1 min readOct 15, 2023

--

I keep reading these articles on OBT and they all claim the same false information -- That dimensional modeling was created to minimize storage and compute costs. While that was a concern, it is only a part of why we use dimensional models.

If we take the simple use case provided in this article and we don't care about any dimension values not mapped to a row in the fact table, and we don't care about the dates when the dimensions were available, and we don't have slowly changing dimensions tied to any business processes, then OBT is a great way to store the data. Why not make it easier for downstream consumers. I'm a fan. I would definitely create OBT on top of the dimensional model anyway. Many BI tools expect it.

Maybe I misunderstand OBT, but how would we answer questions like these:

What products were available to be sold in 2022?

What are the user ids under account X?

What user ids roll up to sub-organization X within organization Y?

etc.

In cases where the dimensions themselves are important, OBT doesn't seem to fit. Like anything, OBT is a tool we should have in our toolbox. It solves a lot of problems, but data doesn't always fit into a single pattern. We need to understand multiple patterns to solve whatever problems come our way.

--

--

Leo Godin

I’m Leo and I love data! Recovering mansplainer, currently working as a lead data engineer at New Relic. BS in computer science and a MS in data