Name: Data modeling interview filters so many data engineers! How to model slowly-changing dimensions
Uploaded: 2026-03-28T09:23:01.000Z
Duration: 2 min 58 s
Description: Thousands of YouTube videos with English-Chinese subtitles! Now you can learn to understand native speakers, expand your vocabulary, and improve your pronunciation...

Transition words & phrases

A lot of data engineers in data engineering interviews get filtered out in the data modeling round because they don't really understand it very well.

Modal verbs - B1

And if you ask around, like, "Oh, how do you learn data modeling?" some people might say, "Go read the Kimbell book." Some people say, "Go read the Inman book," or whatever, right?

Adverbs of manner

And, um, at least for me, the way that I learned data modeling was mostly through experience and, like, through testing things out and trying things out and experimenting while I've been on jobs.

So that's been one of the ways that I've been able to kind of learn efficient data modeling.

So here are a couple key components to efficient data modeling, right?

So one is learning how to model dimensions effectively, right?

So dimensions are interesting because they are, like, attributes of a noun, right?

So maybe it's an attribute of a human, like a user, or maybe it's an attribute of, like, a listing, if it's an Airbnb or something.

And some of these attributes are unchanging, right?

Like, for example, my birthday is unchanging.

It's the same—like, that day I can't change.

But other dimensions for me do change, right?

Like, my age, for example, changes over time, right?

Right now I'm 29, but, you know, next year I'll be 30, right?

And that dimension is a changing dimension for me, right?

Same with things like maybe my favorite food over time changes as I, you know, learn more cultures and I eat more food.

I'm like, "Oh, actually, lasagna is no longer number one," right?

Maybe it's, uh, some sort of curry or some sort of other thing, right?

And so those are—the big thing I'm talking about here is understanding, like, when it's a fixed dimension, right, versus a slowly changing dimension, which is a dimension that kind of changes with time and has that, like, temporal component to it, like my favorite food.

And so then you can kind of learn about the different ways to model slowly changing dimensions, uh, because fixed dimensions are really easy because it's just, like, a column and a user, right, ID, and

then the dimension value, and that doesn't change.

But slowly changing dimensions, there's a—you know, there's type one, type two, type three.

I really like, uh, the type two—SED type two way of modeling.

I think that that's the most effective way.

And how that works, right, is, uh, you can have multiple rows for a dimension,

but each row has a start date and an end date.

And, like, if it's the current one, the end date is just, like, really far into the future.

It might be, like, 9999, like, it's like the year 9000 or whatever, right?

And it's like—so it's like that means it's, like, the current record.

And so that's how you can kind of get the SEDs to be—or, like, slowly changing dimensions to, like, be in a data model that you can easily use for, like, joins and queries and stuff like that.

And you can understand not just, like, what the current value is,

And being able to have that historical context is also very important for, like, backfills and understanding your data sets.

So definitely keep learning more about data modeling.

I'm going to keep posting more videos, you know?

Keep, you know, follow to get more knowledge here.