Jordan Goldmeier argues Excel remains essential in data science despite its bad reputation, especially in finance and enterprise settings. He explains how its visual formulas work for prototyping algorithms, cleaning data via Power Query, and communicating results to non-technical leaders. Excel complements Python and R rather than replacing them.
Chapters
You know, there's a trope in the data space that data scientists will refuse to do anything in Excel, no matter how easy the task is, even if it means just computing the mean of a column. And your show, your book, sorry, shows how Excel can be a very powerful tool for data science. So maybe to start off and set the stage, how relevant or not relevant is Excel for data science?
Well, I think that the way Excel presents itself in data science is continually changing. So we. Much of the critique that came, really, if you think about the history of Excel, first it was like your choice was Excel or Matlab, your choice was Excel or S or Jump or something like that. So it was either you do a lot of work in another coding language or you try to make it work in Excel and it's just your computer's freezing and overheating catches on fire. But sometimes it works, right? So I think what happened was all these new open source technologies came about, so Python and R, and there was still in the executive level leadership, this idea that we always need Excel. So this developed a lot of hatred from data scientists who were forced to use this product, who were like, this is not the right way to do this. And having lived on both sides of that, I think that Excel does slot in very nicely within data science and that many data scientists are at their own expense, not including it in one of the skills that they should offer. So for me, the way Excel, like, assuming that it does continuously evolve, but the way Excel really slots in, as we talk about in this book, is creating algorithms where the components are actually very visual and very seen and they have formulas associated with them so you can really understand the logic and you can walk someone through it. And as well as if you were going to start a data science project and you had to communicate it to leadership, you can come in with all this Python code, you can run a jupyter notebook, and maybe that works for your next level manager because they were you five years ago, but for their manager, their manager's manager, they just want the facts. And they want something they can click on that makes sense. And even if they are programmers, chances are like, I have my own company, I have staff, so I'm a CEO and I tell them all the time, just the facts, right? I only want the facts. And so Excel gives you just the facts if you do it right.
Okay, that's really great. And you know, we're going to definitely uncover a lot of the use cases that could be relevant for data science and Excel. But you know, I Mentioned that trope of data scientists not wanting to use Excel at all. But there's also another kind of really popular meme in the kind of data space of literally Excel carrying around the world's GDP on its back. Right. So, you know, Excel is widely used by a lot of folks. It has quite improved a lot as a tool, maybe. Why do you think Excel still has a bad rap today despite all of these improvements? Kind of expand on that notion a bit.
Sure. So just. I'm going to throw the book up just so we have.
Yeah, it's. It's really great. It has a spreadsheet vibe and everything. Like, the cells are like, totally.
Right.
Yeah.
I always wanted to do this for the COVID of a book and then finally I was able to.
So, yeah, it's really great.
You know, I like to compare the, like, use of excel to the TI83 calculator. I guess it's TI83 plus now. I don't know what the kids are using. All right. Maybe they're not even using that one anymore. Anyone. During those years, when I was a kid and I'd hang out with these tech nerdy people who were coding, they'd have the jobs I want. Like, I'd go to synagogue, ask them a million questions, and I Talked about the TI83 plus and they went on this thing about how Casio makes the better calculator. Right. That may be true. There may be products that do what Excel does better. But the way tech works in our world is it's not about what we idealize, it's about what people are using. So we can debate what, like the degrees that Microsoft made people use Excel or took them away from other people. But as you said, it does run the world's financial system for better or for worse. Like, I can sit here and list all the things that it does well, you can list the things you don't like. It's not like you're wrong, but it doesn't change this fact. So I think if you are an analyst and I. You can send your hate mail to jordananarchy.com there. But you, like, you as a data scientist, like in the career field, the data scientist sits above the analyst on the ladder. But the truth is, you're like, in the broad sense, you're analysts. Okay? This is what you do. Use your brain to analyze this stuff. You need to know Excel. I mean, if a bank, if you work for a big bank, you may be right about all these products and this and that, but over the years, they've been using Excel and so even though I love vba, I like the macro coding language. For those who don't know behind Excel, I wouldn't recommend people use it for most problems, but the entire banking system is run on it and there are cottage industries to approve that. But it wouldn't change the fact that a lot of things that are important are run on it and people don't want to change those things quickly.
Okay, and you mentioned, you know, early in our discussion, you mentioned that there are quite a few use cases where Excel could be very relevant for data scientists. One is where you have to build systems or algorithms, where it's very visual. One is when you have to show a prototype to leaders. Maybe walk us through these use cases in a bit more and walk us through how Excel could potentially work with other popular tools in the data space, such as R and Python as well.
For sure. So I think that it's not like. It's not like that. I'm an Excel maximalist on this. Certain things you have to do, even things that we do in the book, like, like the book is more of a textbook to explain all these things. But I would not. We built like a spam detector using the bag of words in a Bayes and naive bayes. Okay, so to classify. I wouldn't do that in Excel. I mean, that doesn't. It's not deployable in Excel. But if you wanted to explain to someone how it worked as a professor, as an analyst, as a manager, Excel is great. And if you didn't want to run the whole entire training set, but rather just needed to explain to someone using a small amount of data, Excel is great for that. Again, you can't put that in production, but it's good for that. The other thing Excel is great for is data cleaning. If you've used DPLYR and R or you've used Pandas and Python, it's very similar. But if you had to explain it to someone again, what you can do in Power Query is actually much quicker than you can do with dplyr. And I love dplyr and I can crank it out, but again, it makes it a little bit easier to understand what's going on. I can, for my own work use R, but for my clients, I'm going to use Excel because I'm talking to people who are finance people, they're not coders, but they understand what Power Query does. And then I'm trying to think here because it's not really about what is right now, right now. In the world of data science, all these other technologies are the better choice. But there will be a point in the future where you could plop something in Excel, send it off to be worked on by some algorithm in the cloud, and get the results very quickly right back. The same with Power Bi as well. Again, it's good to know it now because that's our future. Because the difference between these tools and like, yeah, the difference between these tools, like, right now is very obvious, but I think that we're going to reach a point where they converge.
