What is Descriptive Analytics
Welcome back! On this weeks episode of Data Science Wednesaday, Decisive Data's Lead Data Scientist Tessa Jones takes us through a basic understanding of what descriptive analytics really is. Check back next Wednesday for an all new DSW video!
Welcome back to Data Science Wednesday. My name is Tessa, and I'm a data scientist over here at Decisive Data. And today we're gonna be talking about descriptive analytics. You might be asking, "Why are we talking about descriptive analytics? That's not data science." I know it's not data science, right? When we think of data science, we think of predictive, prescriptive, diagnostic, but not really descriptive. Having said that, descriptive is really necessary as a launching pad. Before we can really dive into the advanced analytics, we want to have a really good grasp on what's going on.
So, as we walk through an example I want you to imagine that you are a grocery store owner, and you want to optimize your shelf spacing. So if you have products on your shelf that are not selling very well but they're taking up a lot of space, that can have a really negative impact on your revenue. So you want to optimize that. So, your questions are, what are our top selling products? Also, it's what are our least top selling products? So you kinda wanna know both. So let's drive through an example.
So when you're looking at descriptive analytics, most of the time you're gonna be looking at a dashboard through Tableau, or Power B.I, or something of that nature. and you're going to drive through to try...drill through to try and answer these questions. So let's look at this. We look at the upper left-hand corner and we're also looking at revenue by product category. So we're saying pantry, meats, produce, and you see that your pantry products are doing really well. But you wanna know, well, which ones are doing really well, and which one they're not doing so well, so that you can, you know, optimize your shelving.
So a really good dashboard would be interactive, so that when you click in here, you can go over here and see how good each product in that category is doing, right? So then you're over here and you see that Frosted Flakes is doing really well, Captain Crunch is doing really well, and Joe O's is not doing very well. So then you can go and you can have a conversation with your inventory people or your executives and you can say, "We need to buy more of this and less of this," to optimize your revenue. It can really drive really good conversations.
A lot of times people who are using these dashboards don't really understand everything that it takes to build and support this kind of a dashboard. So let's just take a minute to kinda talk about that. We're gonna break it down into four basic things. Cleaning, relating, summarizing, and visualizing. So cleaning, we've all kind of heard of this, right, before, dirty data. Nobody really wants to deal with dirty data, because you might have text fields with, some have caps, some have lowercase. You don't know how to relate them to each other. You have dates that don't make any sense or nulls that are not handled properly.
All of these kind of things can cause inaccurate reporting, which you know you don't wanna deal with. So then the next thing down is relating. So a dashboard like this could be built on one table, it could be built on hundreds of tables. But you need to know how all of that data is related. For example, if you have Captain Crunch up here, you need to know that Captain Crunch belongs to the pantry category. And the only way that you do that is by relating those different pieces of data together. So that's a really crucial step. Next, you wanna summarize your data. So this is really how you get an overarching view of what's going on.
You don't wanna look at your data line by line. You don't want to know what's going on day to day per se, but you want to be able to summarize. So you maybe want to know well, how many sales happened throughout the whole month or by category, or into different areas. Any kind of different thing is gonna involve some sort of summarization. Next, this is a pretty crucial one, visualization.
So as we talked about, you know, this is the visualization here. And so when you're creating this visualization, that's really where the geeky data side of everything collides with the business decision-making side of everything. And so it's really critical that this not only is very usable for the business person but also displays everything accurately. So that's a pretty crucial step.
So what is descriptive analytics? Well, we know that it's not data science. But how do you know when you're in descriptive analytics? Well, anytime you're answering questions that describe your business, you're probably in descriptive analytics. And what does it take to support descriptive analytics? Well, it takes a lot of data massaging, cleaning, relating, getting everything kind of prepped, and then creating these really great visualizations. And that's when you know that you're in descriptive analytics. Thank you.
Posted by Gage Peake