Here you'll find the recordings of our Techspert Talks sessions (formally Support Techspert Thursdays).
This session addresses:
- Deep dive into visualizations
- Demonstration of most effective charts
- How to exploit controls to the max
00:00 - Intro
01:36 - Decision Tree with Election data
03:33 - Radial Dendogram - map chart
04:34 - Changing map chart perspectives
05:45 - Cartograms
07:14 - Dynamic labels
11:01 - Tracking Qlik Sense features
12:00 - Time Plan calendar - Trellis table
14:20 - A and B Testing
16:28 - Custom Time axis
19:04 - Custom Time with colors
20:12 - Less is more
21:26 - Q & A
Hello everyone and welcome to the May edition of Support Techspert Thursdays. Today's presentation is Top 10 Visualization Tips for Qlik Sense Enterprise with Patric Nordström. Patric would you kindly tell us a little bit about yourself?
Thanks Troy. It's good to be here. I’ve been with Qlik since 2017. I came aboard with Qlik with the Idevio acquisition. I’ve been working in the Qlik echo system since 2007 or something. I’m the product manager for Qlik Sense Visualizations and Qlik GeoAnalytics.
So, that's what I do.
Great, thanks. And what will you be sharing with us today?
Most of my time I spend talking to customers, listening to their problems and suggesting different solutions; coming up with requirements for R&D. In many cases, the customer doesn't know the product that well so. There's so many things that you actually can do with the standard settings inside of Qlik Sense; so I see it as a bit of a mission to remind everybody what a great tool it is and how many things that we can actually do within Qlik Sense without having to build something very special, because it's so flexible.
Yeah, I know you have a lot of wonderful tips here. I don't think we'll go through them all, but I thought we'd start. Since being an American, I carefully watched the presidential election recently, and trying to explain the intricacies of how that works can be challenging. And I understand you have a solution for that?
Well, I was playing around with decision trees lately. It's useful in many different applications and one of them is for visualizing the outcome of elections.
I was looking for some public data to play around with and the election is of course quite fun to play with.
We have a couple of charts that are suitable for this type of data. One of them is the Org Chart. It's just a table with nodes and information to the parents. The parent would be the Node ID; just a table linked to itself.
And that's a default out-of-the-box diagram, right?
Yeah, so it comes in the Visualization Bundle. So, it's over here in Custom Objects > Visualization Bundle.
Okay, there it is.
So, it's meant to visualize company rosters, but it can easily be adapted to show the outcome of decisions. This tree shows the possible outcomes of the swing states.
Okay, and how would you interact with this?
Yeah, so this is how we start. And then we decide whether we think the Democrats or the Republicans will win in Florida. So, if we think the Republicans will win, we will come to the next swing state which is Pennsylvania, and we can see the current state. If Florida is won by Republicans, Pennsylvania's by Democrats and then the Democrats win Ohio, and the Democrats wins. So, it's a nice way to show a rag tree, not completely balanced. As soon as any of the parties win, it will actually show that in a darker color. Of course, the Republicans had a different position to start with, so they have to go through a number of elections before they call it a victory.
So, it's a nice way to traverse a tree. It doesn't show the whole tree at the same time. If you're more interested in the complete decision tree, you can instead show them in a Radial Denogram instead, which is the same type of tree, but I’m now showing all the nodes at the same time.
The Org Chart showed the beginning of the tree in the center, but this one is expanded to show all the branches in the tree. You can highlight the possible wins for the number of ways the election can go, well estimated to go. And you can also highlight when it can be a tie; which is quite…
…interesting that it actually can be a tie.
The last was the Org Chart. What's this it in this case?
I’m using the Map Chart instead.
And this could be good for displaying any kind of decision tree, right?
Of course. I mean you could use this for any type of decision tree data where you have some kind of pattern, where you have a yes or no decision. One outcome leads to another, and you want to show the different types of possibilities.
I’m just playing around with radial coordinates, but you could also use the same tree and have a different coordinate calculations, and then you would have a flat tree; and then you could play around with it if you want to have different types of perspectives: up or down, or left to right. It's just the transformation of the coordinates.
I love that putting it on a cyclical group, so you can switch quickly between different visualizations, and see the same data shown different ways. That's cool.
Yeah, it saves space as well. I mean most of the time, you have many ideas when you want to build a dashboard. And the first thing that you do is that you try to cramp into nine different graphs and diagrams and tables and everything at once. You win in both how the user experience the app, and how well you can perceive your message by showing bits and pieces of the information at a time; not revealing everything at once.
I know a lot of times especially with data from the election, people are trying to apply it to a map. But do you have a way of focusing more on the data points, but still having some sort of geographical elements to it?
Yeah, so that is also something that I’ve tried to build out, showing some kind of simplistic way to show geography.
We do have a Map Chart and you can see the geography is exactly how it is; but you want to show KPIs that doesn't have that much to do with geometry.
So, this is a way to present a simplistic way to show maps as squares or hexagons or circles. In many cases, you have to do this by hand. Instead of doing that, I used an Excel spreadsheet as a template. So, I organized the states in a row-column order, just to place them and then I can move them around. It's not exactly how the Earth looks like.
It serves its purpose to show some kind of simplistic map over the U.S. instead.
Exactly. It's kind of elegant in its simplicity, I like that.
Yeah sometimes, it's the actual geometry gets in the way. And by modifying this grid, you want to do it for other countries, you can just change the names of the states for the region that you're interested in and reload it. And you will have this type of cartogram also for your special region.
That's great. Now, you're obviously an expert on charts and visualizations. In your experience, when users first notice a diagram, what is it that stands out to them? What do they focus on?
Yeah, so that is another thing that I usually bring up all the time is that: the first thing that they look at is the title and the subtitles. So, they spend more time reading that instead of reading the plot, which is quite surprising to us; that they are so focused on the actual charts. You don't really know what the user will make out of the dashboard, but I think we can come up with better titles. I mean the easiest way to do it is just to have the dimensions showing ‘Sales by Year,’ say ‘Shares by Category,’ some kind of explanation. But you can actually do it do it better in my opinion. If you look at the chart, what am I trying to communicate? If it's a line chart showing sales over time, then perhaps I’m interested in the trend. And then it makes sense to show how big was the increase or decrease in the last period. And we can also show the total increase over time. You could do that by building up the titles dynamically using Set Analysis perhaps together with Maximum and Rank to build up the titles. Then the titles will be dynamic. Even if the user is making a different type of selection, or the data is reloaded; the titles will still be valid. We do have a Snapshot functionality with the Storytelling, where you can add Annotations to the chart.
But this is a way to build some kind of dynamic, that will be true whatever kind of data you throw at it. And you could do this for the Line Chart, but you could also do it for the Pie Chart. If Pie Chart is probably something about shares, so you probably want to highlight the largest share, and then you could have the second and the third. Bar charts is mostly about comparisons. So, then you could have the title express how big is the largest value compared to the second one, or you could also incorporate what the top three represent in the total value.
That's a much more powerful title, get a lot more information immediately from the chart than just the default.
Yeah. It's an eye-opener at least to me, when I when I played around with it, that it actually carries much more information that.
I’m lazy at this as well. I usually said just some boring kind of titles, but it really pays off to make it a bit more elegant or try to use that space that you have.
It's really well done. You mentioned: you built these dynamically using a Set Analysis. Is it possible to take a look at how those are built?
Absolutely. It's not like rocket science. If titles is switched on, I just picked the maximum year divided by the year prior, and then I could see whether it goes up and down. It's not that that hard; especially if you know a bit more how the dashboard should be used. Then you could probably, you could tweak it even further. In this case, I’m also using the new function of the Dimension-based Reference Lines. So, I throw in a little Annotation inside of the chart, and that's goes over here.
I picked the last year like that. Just to place it somewhere, and then I add the title the increase or decrease within the, in the label. And then I have a show condition, in this case to switch it on and off to make the point.
Is that, do you know when that feature became available?
So, that was one of the latest editions.
So, okay. I know that's a common problem that comes into Support actually. We get questions all the time about what features are available in what version, especially when it comes to on-prem customers wondering when they should, to what version they should upgrade. So you have a way of keeping track of that?
Yeah, I have to put it in an app; but otherwise I can't keep track of that.
So, what are we looking at here?
I’m keeping track of all the releases in or all the new features in each release that are relevant to my, to my area.
Dimension-based Reference: already in 2020 September; so it's been around for quite some time.
Is this, is that app publicly available?
It's not, but maybe it's something that we should have put in, because there's a lot of people asking for it.
But I tend to meet more and more cloud customers. And they do already have all the functionality. They don't worry so much about which version they're on. I like to keep it also to show the progress, that we are actually adding a lot of functionality to the product all the time.
Do you have a favorite kind of chart, or one that you've used a lot recently?
Yeah. I can't help myself. I was planning a building project at home. You have to apply for the building permit; you have to get the permit; you have to wait a while before it becomes in effect; and then you have to order the stuff that you want for the building. I decided I want to show it in some kind of calendar, with each one of the months together, with the weeks. And then I added some sliders. If it takes shorter time for me to get a handler, and maybe they are quicker to give me the approval. In many cases, you see Gantt charts for project planning. But in this case, this was for our community. I wanted to talk to people that didn't perhaps play around in dashboards that much. So, they were more comfortable showing data in a calendar, because most people have seen a calendar.
Exactly. But what kind of object are we looking at here?
This is the Trellis Container.
Instead of having to make seven tables, I can just make one table and build out the functionality inside of it. I have added the days, then the weekdays. Whether it's Monday, Tuesday, Wednesday; and some background color to show if it's a weekend. I added my activity last, the week number and then I have four variables that adjust the timing for things. It doesn't take that much data either. The only thing that I placed was the variables, and then I built up a calendar from the start date to the end date. I’m just auto generating all the dates between max and minimum date.
That's really cool, actually. Just be able to quickly see how it all these different decisions and times will affect your end date.
Yeah, it becomes really evident. People are asking: when do you think we will actually have this in place? And then it's an easy way to show people, understand that this thing has to be in place before this one, and there's a certain amount of delay time before we can actually have it in place.
A common challenge in my department at least one that we face is testing new changes in a system to kind of analyze if there might be improvements and where. How would you recommend visualizing that for that?
A and B testing is probably the most popular analysis that you could do.
Of course you could use Qlik Sense both to calculate the statistics behind the A&B test, but also visualizing how well you can rely on the data. You have two groups: one where you let them show the difference or the change; and the other one keeps the old. And then you try to see if you have some kind of confidence that the changes that you made really made an effect on. In this case, it's conversion. So, I like to show it as Bell Curves. So, this is the conversion rate for the control group; and this is the conversion rate for the variance group. And then just calculating conversions between visitors. There's a conversion rate uplift is the percentage of how much better is this in the conversions; and then you can calculate the standard deviation. Said value, P-value and in this case. It's all about the P-value, so if the P-value is below 0.05 then you would have a 95% confidence over significant test results. I added a slider where I can change the conversions in the variance group. So, if I increase it, I have higher confidence; if I decrease it, so few other conversions, I can see that now it's no longer statistically safe, or I can't be sure within 95% that the change is a significant test result.
And that Bell Curve is kind of a powerful visualization as well.
And it's quite easy to build up in Sense as well. I mean you already have the normal distribution as a function in the engine. To me, it resonates well. Of course, you could show it just as a table. I like the way that it spread out the estimate of the normal distribution.
I know in Sweden when you're looking at calendars, and stuff like we were before; weeks is very common. Do you have a way of making custom times, like with weeks instead?
When you drop an Excel file inside of Sense, and you're loading it. It will actually create a master calendar for you, and it has some basic settings. But it maybe it's not that super well known that these time-aware charts, when you have a continuous axis, you can actually decide how they should look like. The normal standard way is probably to show years, and then months, and then dates, perhaps. But if you work a lot with weeks then it makes sense to show week numbers instead; which is quite popular in in Sweden. Not so much in the U.S., perhaps.
When you build up the calendar, you declare derived fields (so to speak), and then you have expressions for the different parts of your time axis. The meta information that goes in (that decides how the time aware chart should look like) is the way that it's tagged. So, dollar axis (that will actually show on the axis) and then you have two different ways to show qualified: that means that this axis will show when you are zooming in. And Simplify: this is the axis that will be shown when you're zoomed out. And you could also have helper expressions that will not be shown in the UI, but you can if you want to build up something more complex. So, now I’m sort of zoomed out. I show the years, and then I show the week number for the first week in each quarter. I can't have all 52 weeks, because then it won't show me any week numbers. As I’m zoom in, you can see the individual weeks. And at a certain time, it will switch over and show me the individual week numbers instead. And if I go even further, it will actually show me the dates again.
Oh, wow! Even all the way to times. That was pretty cool.
Yeah, as we switch over to week numbers, so we show all the week numbers. We also change the quarters to show both the year and the quarter number. Otherwise perhaps, you get too lost when you are looking at week 27, whether you are on 2020 or 2021.
That's good to have that additional reference.
You can build up your own fiscal year, perhaps. In Sweden it’s quite common to have fiscal years that start at a different month. Could have your fiscal year starting on May 1 for instance. But there is actually some customers that have a fiscal year that starts at a certain week number. Then it becomes a bit tricky, because if you have a year that started week 36; then it's no longer symmetric. You will have years that are different length depending on, because week numbers are kind of sliding. It's easier to calculate or build up an offset if you start on a month rather than a week. But you could still build something. In this case, I’m playing around and using colors instead. This is a customer of ours that have a fiscal year that started week 36. The week numbers are correct. So if I zoom in here, we can see that they actually starts at one. But it's actually week 36 in this case.
And I’m using the color functionality together with the time axis to show where the fiscal year starts and ends.
Now I know you have a lot of tips and recommendations there in this app that you put together and developed. We don't have time to go through them all of course. Do you have any general advice for developers who’re trying to decide what is the best solution to the problems they're trying to solve?
Yeah. In many cases with design or anything, that you usually try to add too much stuff. And you're not done when you have no space to add things. It's probably when you should start to take away things that are less important in order to highlight something. So, you're not done until you have taken away everything that is not relevant any longer. And especially if you work with electronic dashboards and not with paper reports; then you have all the possibilities of adding interactivity that can fill the gap. And the human brain usually works really if provided an overview first; and then you get details later. So, you could use when you hover over something, you can get a custom tool tips that provide additional information for that space that you're interested in. Of course, the whole paradigm of Qlik Sense: with making selections and narrowing down the scope that you're interested in, also helps a lot in order to consume a complex data set.
Great! Thank you very much. Okay, now it's time for Q and A. Go ahead and submit your questions to the Q and A panel on the left side of your On24 console. Patric, which question would you like to address first?
So, we can go from the top.
What's the most effective way to display filters? Especially if there is a number of them are used to struggle to understand on how to use them?
I would say that the most common used filter is the Selection Bar. So, when they started to use some kind of selections that is the most peoples are interacting with, then you have the possibility of adding filters to your to your page, or to your dashboard to help the user for an initial selection. And that's what filter pane is for. I’m not sure if everybody knows how the filter pane works. Because if I’m adding something to the filter pane, I can add additional fields to it. If I add a lot of different ones, it will actually collapse and help you to keep them organized. If you want to make selections on it, you can still do it; but you can still also you can expand it. So they can get easier full screen view of all the selections. I like to work with the titles with the filter panes. Instead of just showing the title for the field, you can actually (like I did here in my feature overview). So, instead of just showing Year here. I’m showing: if I make a selection, it will actually show me that in the filter pane. Of course, this is also included in the selection bar; but if I make another selection, I have a dynamic title, so instead of showing Year, I’m showing the actual values for the year. It works really well if it's something obvious; like years, because you don't have to spell out that this is a year. Most people recognize that directly. And you can do that with other things as well, so.
This is 2021 button, and canon, and you could have a look at it that at the same time. So, that's my best tip for selections. I also, if you're making some kind of guided application where you want the user to click and make a certain type of selection; you can help the user to have - you have buttons that you that the user should hit. And you could apply either a selection directly, or a bookmark.
Instead of showing everything. There's also the Global Selector. That will show you all the fields which; I don't know how many users are actually using, but this is a quick way to get to see all the data that is available within your side of your app.
All right. Next question?
Yeah. Next one is: can we evaluate which visualizations or metrics takes the most time to load?
So, in my experience, when you visualize something, it's much about the actual metrics behind. So, the expressions that you make, that is what can actually take time to render. So, I usually try out all my expressions first, before I, or especially if they take a lot of time. I try to debug and try to write better expressions just by looking at them populating a table, perhaps to start with. And know that they are actually producing the result that I’m interested in.
Right. That makes sense. Testing is always important.
Yeah, I mean especially, you probably start with some kind of subset, and then you load the full data. And it might be behaving a completely different way. So, it's, you have to work with the app through the different phases I think.
Okay, next question?
Where can we find more information about Qlik Sense extensions and maps?
So, it's all on the Help Site. It's organized with the different products, so I would say that if you're looking for reference information, then you will have all the information at the Help Site. Then there's additional information; perhaps more tutorial specific ones. The one thing is the reference documentation, but you sometimes you would need, you want to do something very specific. And then reference documentation might not take you all the way. So, there's the Qlik Design Blog of course. And then at the Qlik Community. We have different forums. So, if you're going to perhaps for Qlik GeoAnalytics, so.
I notice you're very active there.
Yeah, I try to spend a lot of time here. So, then there's a tutorial section where we could see the in the document section for GeoAnalytics, where you can find tutorial built out tutorials if you want to load a shapefile for instance or if you want to do something more specific.
Okay, next question?
Can you give an outlook which functionalities from the on-prem version will follow in the cloud edition? Which gaps will remain?
So, most of the things we release in the cloud version first nowadays. And then we gather all the new functionalities, and they will go into the on-prem version four times a year. So, if you have a cloud account, it's good to keep track of all the new features; because you get them immediately in Qlik SaaS, so that is a good way to keep track of new things.
When are you going to have out-of-box colors font borders, etc.?
So, you might have seen that we have added a lot of features. I think we pushed up more than 96 different features the last five milestones. So, there's a lot of - we focus on three different areas. We push out new charts, completely new charts, new functions inside of the charts, more like accumulations and trend lines, etc. And then we are also adding custom settings. So, in the Line Chart for instance, you could have dashed or solid lines, you could have curved or straight lines. So, that's type of short specific settings. We will also add more common settings to, like that are more common for the short background; perhaps the title. So, that is something we want to add, but and we will add it continuously, released by release. But we wanted to start out with the most important parts first. But there will be more settings like that as well. It's relevant (I guess) for people who want to have more control over their application; and want to build something very specific. You can control these settings already today by using a Custom Theme. So many of these are exposed that way you could also use css to tweak.
And if you want complete freedom to do anything that you want; perhaps you should also - you should build a mashup instead. Leave the Sense application and just embed the charts that you like, and you then you can tweak it any way you see fit.
Okay, Patric we have time for one last question.
We can go: which is the best way to connect with to see more of these type of visualizations?
I would say to stay tuned to the Qlik design blog. There's a lot of nice articles posted over the years on how to design and how to build for this specific app or for. This app with tips, it's part of a series. So, this was part number five. So, there's four other apps that you could, they're all available on the Qlik Community. So, if you search for in Qlik Community and “Top 10,” you probably get a hit for this one. And they are linked together, so you could explore them. The Qlik Community is a great resource for sharing and for finding inspiration to build things with Qlik Sense.
That's really great, because I’m sure a common question amongst our attendees is ‘where can we get a closer look at this app that you've been sharing?’ Great. Well, thank you very much Patric! I think this will be helpful for a lot of people.
Yeah, thanks a lot for having me here. It's great to be able to spread the word on how things that you could do with Qlik Sense.
Okay, great. Thank you everyone. We hope you enjoyed this session. And thank you to Patric for presenting. We appreciate getting experts like Patric to share with us. Here is our legal disclaimer. Thank you once again and have a great rest of your day