Ryan discusses the spectrum of analytics you can do in Tableau, his observations on the current state of the practice of analytics, how these terms help align stakeholders, and most importantly, how each help derive value from the data.
Hi, this is Ryan with Playfair Data TV. And in this video, I’m going to be describing the four different types of analytics that you can do within Tableau Desktop. Those four different types of analytics are discovery, descriptive, prescriptive, and predictive analytics.
I’m going to show you some examples of each of these over in a dashboard I call the Super Sample Superstore dashboard. And if you haven’t seen this, you should Google it, or look down on the related content under this video. I’ll link to it. It’s just a good way to teach strategy, and there’s a lot of different examples within this dashboard.
But we’re going to start with discovery analytics. This is the type of analytics where you don’t necessarily need to know what you’re looking for. You’re just dragging stuff onto the view and seeing what emerges. To represent this type of analytics, I’ve got this blank slate.
The reason I’ve got that is, again, you don’t necessarily know what you’re looking for. This is just you at your desk as an analyst kind of exploring the data set, connecting to something new, dragging some fields on the view, see if something pops up. I’ve seen some huge wins with this type of analytics.
No, your colleagues may never see most of what you’re doing at this phase of the analytics. But what I find is, once you go down a certain path and you’ve committed to what you’re going to look for in a specific dashboard, you kind of get those horse blinders on. And you just start to measure those two or three measures across those three or four dimensions. And you’re stuck looking at the same things day in, day out.
You might be missing the forest for the trees. There might be some other measure or trend that you should be aware of. And discovery analytics provides that flexibility to find something new. So there’s a lot of value in this, even though your colleagues may never see this aspect of your analytics. But this could take up even 50% of what you’re doing.
The first tangible type of analytics that I have to share with you I call descriptive analytics. I don’t call it that. I definitely did not come up with this. But it’s called descriptive analytics, because it’s describing what happened in the business.
I’m a really big fan of the 80-20 rule. If you’re not familiar with that, it says that 80% of outcomes are due to roughly 20% of effects. But it’s just amazing how many things in business and in life tend to fall into that ratio, 80-20.
And one of my observations has been that roughly 80% of companies are stuck in a spreadsheet mentality. They’re doing 100% of their analysis in Excel-like spreadsheets. Of the 20% that have bought in and they believe in the power of data visualization and interactive analytics tools like Tableau, I would say that 80% of them are stuck at descriptive analytics.
It’s a huge leap forward compared to a raw spreadsheet of data. But there’s still some limitations to this, which is why I’m going to describe the last two types of analytics. But again, it’s called descriptive, because it’s describing what happened in the business.
One of my objectives with this Super Sample Superstore dashboard was to make one kind of cool corporate example out of the Sample Superstore data set that we use for most of the data examples here at Playfair Data TV. And there is one big insight that’s standing out for me from this sample data. And it’s related to Profit Ratio.
So this dashboard is looking at three KPIs: Sales, Profit Ratio, and average Days to Ship. And Profit Ratio is down pretty significantly. I can tell that by these gauges in the middle.
Those are called bullet graphs. They’re similar to bar charts, but that line provides some extra context. Usually, that’s some type of comparison. Usually, it’s either a prior period or a goal if the business has those defined, which isn’t always the case. By the way, we do have a video how to make bullet graphs here at Playfair Data TV if you’re interested in making this chart type.
But this one really stands out to me. It’s showing profit ratio’s down. It looks like the biggest gap between the current performance and the comparison performance. If I hover over that bullet graph, I can see on the tooltip there that the period-over-period change was almost 4 percentage points.
So that’s a really interesting insight. That would’ve been very hard to find in a million-row spreadsheet. So again, descriptive analytics is a huge leap forward. But the limitation here is I don’t know why that profit ratio was down 4 percentage points. And I also don’t know what to do about it.
So I find that, when people get stuck at descriptive analytics, they kind of very quickly fall out of love with these types of dashboards. Because they become stale very quickly. I’ve seen the insight, but this dashboard is really only an alert-style dashboard.
It told me something happened. But I have no idea where to look next. I don’t know why that happened and what I’m supposed to do about it.
That’s where prescriptive analytics comes in. And that’s why I have a tab for this. In my opinion, you should be aiming for prescriptive analytics. I believe that this is the most valuable type of analytics where the analytics industry is currently.
We are going to talk about predictive analytics, one more type. But to me, this is the sweet spot. This is what you should be aiming for.
Let me show you how we’re going to take a descriptive insight, and quickly figure out why it happens. We’re going to diagnose the issue, and then prescribe something to do about it– which, by the way, is where this name comes from. We’re prescribing ways to fix something in the business.
The first thing I’m going to focus on is the Profit Ratio KPI. I’ve got this dropdown that allows my end user to choose which KPI becomes the focus of the dashboard. Right now, we care about the Profit Ratio KPI, or Key Performance Indicator. So I’m going to click on that. And that becomes the focus of the dashboard.
There’s also another video here about how to allow your end user to choose what dimensions and measures are being used on the view if you want to check out that other video. As soon as I made that switch, I start to see a story emerge. These states are colored by the period-over-period difference. So that bright red state, which is the state of Ohio, that’s indicating that the state of Ohio had the largest period-over-period decline. So that’s a piece of the story.
But I’m going to go deeper on this analysis, and also, change my scatter plot here on the right. Instead of Sales as the y-axis, I’m using the same tactic which allows my end user to choose which measures are drawing this scatter plot. And I can change Sales to Profit Ratio to look at this in a different lens with a different chart here.
And more of the story is emerging now. I can see that the Tables sub-category is actually the only sub-category that had a negative profit ratio during the period that’s selected on this dashboard. And because I left the Discount metric as the x-axis by chance, I can actually see a pretty solid reason I would expect that Tables was unprofitable.
It looks like it was the most discounted sub-category. Out of all 17 sub-categories, it looks like it was discounted almost 40%. So a story is really starting to emerge here.
Profit ratio is down in the state of Ohio. And it looks like the Tables sub-category is the main culprit. And it could very well be because we are discounting that too heavily.
If I’m an analyst, I’ve not only diagnosed why something happened, but I probably have a pretty good idea about what to do about that to alleviate the issue. I probably need to do something in Ohio. And I probably need to discontinue that discount– or at least, reduce that discount– so that my tables get back to profitability.
Because I am so big on this type of analytics, I’ve come up with another little hack that we’ll discuss in a different video. But it allows the analyst to put their own insights and recommendations in line with a Tableau dashboard. That’s why I’ve got a third tab here called Annotations.
This leverages two parameters for each insight. The first parameter has a data type of String and allowable values of All. When you set up a parameter in that way, this essentially just becomes an open text box.
So I could just type in my insight. So profit ratio is down significantly. We’ll say, reach out to the manager in Ohio and discontinue the discount on tables. Click Enter. The second parameter for each insight allows you to determine whether it is a positive, neutral, or negative insight. And based on your selection, the circle next to that insight will be color-coded: blue for positive, gray for neutral, red for negative.
This one was negative. So I’ll click Negative. And you can get a preview of your insight at the bottom. Now if I go back to my Prescriptive dashboard, there’s my insight, as well as recommendation about what to do because of that insight right there in line with the Tableau dashboard.
This will work when we publish it to Tableau Public, Tableau Online or Tableau Server. It will even persist. If somebody on Tableau Server goes to the Annotations tab, types in a new insight, and then goes back to the Prescriptive tab, it will all work right there on Tableau Server. So I really like that little extra tactic.
But to review how far we’ve come to this point with these different types of analytics, 80% of companies roughly, in my observation, are stuck in Excel. That’s the way they’ve always been looking at their data. They haven’t even moved on to making a single bar chart or line graph.
Of the 20% that have bought in to visualizing data and the benefits that come with pre-attentive attributes like color, 80% of those folks are stuck in descriptive analytics. It’s a huge leap forward. We’re able to see these alerts that’s something– there’s something we should pay attention to in a business.
But what I believe we should be aiming for is prescriptive analytics, explaining why something is the way it is in the business and then prescribing something to do about it. And because I am so big on this type of analytics, I’ve got this extra little hack that allows me to put my own insights and recommendations right here in line with a Tableau dashboard. This is pretty much as good as it gets, in my opinion, in the current state of analytics.
If you can’t cause an action or a positive change in the business based on this view, with data that you trust, and just laying out the insight and the recommendation about what to do about it, you probably have a bigger issue at your job happening. There’s probably a bigger cultural thing that needs to be addressed. Because if you can’t cause an action with this, you probably can’t cause an action with anything.
But there is a fourth type of analytics. It’s called predictive. Just to give you one example of how this is on Tableau’s radar, I’ll throw together a quick line graph that looks at Sales by continuous month of Order Date.
I’ll choose Month with the green icon. I’ll jump over here to the Analytics pane. And I will throw a forecast on the view.
So predictive analytics– you’ve likely heard of this. Data science is becoming very popular these days. This is when you’re using things like predictive modeling to predict data points into the future.
I’m a big fan of this. I do believe that this is the future for analytics. But I just want to point out that not everybody has to be a data scientist. You don’t have to know how to do advanced machine learning algorithms. Like I said, 80% of companies haven’t even made a bar chart or a line graph yet.
I believe that the sweet spot currently is prescriptive analytics. That’s what I think you should be aiming for first. I think that’s where most of the value is. But as we get smarter with machine learning algorithms and other types of advanced data science, I do think predictive is the future.
But those are the four different types of analytics that you can do in Tableau. I always try to talk in these terms. I think it helps me align with my stakeholders, and explain my decision making in Tableau.
This has been Ryan with Playfair Data TV – thanks for watching!