The Datalion Blog

Data visualization: Six tips to build a perfect dashboard

Dashboard Oldtimer

In the age of big data, business intelligence and artificial intelligence, data plays an increasingly important role for companies. Dashboards are now widely used, bringing data to the point and thus becoming an important basis for decision-making.

But the optimal visualization of data still is a challenge. With the following six tips you will be able to build a great dashboard and save time.

1. Have a clear goal in mind

What sounds trivial is a common reason for overburdened and meaningless dashboards. First of all, ask yourself the following questions:

  • What do you want to achieve with your dashboard?
  • Who is the target audience for the dashboard?
  • What data do you need to display to get your target audience attracted?
Beispiel für KPI-Chart

It makes a huge difference whether you create a dashboard for your finance department, consumer research or marketing team. Although the database may be the same, users’ interests and priorities can vary a lot. Market researchers usually want details and with many drill-down options, whereas marketing colleagues often have more interest in central KPIs. A dashboard for the management board usually needs a lower level of detail than a dashboard for special departments.

It’s better to create different versions of a dashboard instead of trying to make everyone happy with one view. In modern dashboard tools you can quickly and easily adapt existing visualizations and then offer different versions separately.

2. Create stories with your data

Data and results have a special impact when told as a story. Focus on the questions your target audience will have and visualize the data in a logical and thrilling order. A dashboard needs a leitmotif. The order of charts should be based on the interest of the target group.

How should the user proceed from result A to result B? What data must be shown first to set the context for the following data? Storytelling within reports and dashboards is a very powerful way to get your target audience attracted. Further useful may be e.g. appropriate headings, interpretations and explanations. Pictures can help to clarify your story.

3. Focus, focus and focus

A very common mistake derives from the idea to provide as much data and results as possible. Overloaded dashboards with unattractive information for the target person, have no right to exist. The added value of a good dashboard lies precisely in bringing important information to the point. Simplicity and clarity are king. And if the user wants more information, you may provide him with some drill-down options.

But usually, it’s much more difficult to get things straight to the point than to get lost in details. Unfortunately, this applies to dashboards as well as to all other forms of data visualization.

4. Not only bar plots: choose an appealing design

Example: Aster diagram. This chart type is suitable for a comparison of different brands. In the center a benchmark value can be displayed.

Undoubtedly one of the most important success factors of good dashboards is a clear and appealing design. First of all, it’s about choosing the right visualization: bar charts can be very useful, but there are many other types of graphs to present data. Depending on the topic and objective of the dashboard, it could also make sense to use pictures or text elements. In addition, the corporate identity should be clearly visible.

Don’t worry: you don’t have to reinvent the wheel, there are many templates and examples of good dashboards out there.
Even with dashboards, orient to visual standards that users are already familiar with. Use known icons and navigation elements. A good UX design is key for making the dashboard an important tool in the company. Therefore, test the dashboard to some users before going live. But don’t worry: you don’t need a complex usability test scenario. But go out and watch your target audience during use and take the feedback to improve the dashboard.

5. Let the user play

A major benefit of online dashboards is that they are interactive, not static as PowerPoint reports or Excel spreadsheets. Provide users with the opportunity to delve deeper into data by offering filters or different target groups to choose from. So users can customize the dashboard for their purposes and “play” with data. This interactivity increases the added value of a dashboard and ensures greater use.

Keep in mind, however, that the more possibilities of analysis and customization you provide, the more you have to train users beforehand. Otherwise there is a risk of mis-interpretation. Finding the right balance between too much and too little data in a dashboard is an art.

6. Choose the right dashboard tool

Using a good dashboard software the key factors mentioned above should be implemented easily. But there are also some technical requirements to consider. Of course, the dashboard should also work on a smartphone or tablet device. And ideally in all common browser versions. Since data volumes are getting bigger and the time of your target group ever shorter, loading times must also be as short as possible. Flexibility in data sources is another feature that can decide between success and failure.

It is also good if a dashboard offers export interfaces to Powerpoint and Excel, so that results can be quickly compiled for Executive Board / offline reports as needed. And last but not least: most data is sensitive. Data security should be guaranteed and data protection rules have to apply.

Datalion’s dashboard solution provides the perfect tool to build good and successful dashboards.

DataLion 1.9 released

Experience more lion power – Embrace our new features!

There’s exciting news from DataLion: we released version 1.9 which includes numerous new features we’ve developed in close cooperation with clients & users. We have incorporated your feedback and suggestions gathered in the past months. Among new features like visualizations and export functions, we have also implemented several optimizations (e.g. loading time) and bug fixes.

In short, DataLion is faster, more flexible and more productive than ever. We hope you will love it!

Let’s have a closer look at some highlights:

Yes, more chart types!

Stacked column chart with option to display the mean or the sample size (Example):

Stacked bar chart with mean values

Extended donut chart e. g. for the comparison of Key Performance Indicators (Example):

Nested donut with arrows

New exports, New interfaces! Let’s check out the new export functions

Export of a dashboard or a report with more than one dashboard and charts as PowerPoint or Excel document

Export options

Export of an excel report with multiple tabs in an individually defined format:

Excel report exported with DataLion

Overview of our new features

Data Visualization

  • New chart type: stacked column chart with means and case numbers
  • New chart type: extended donut chart with means e.g. to compare KPI Values
  • Sort charts by mean value
  • Reverse order of stacked column chart segments
  • Top5, Top10 etc. for time series charts
  • Hide empty values and categories in time series charts
  • Separate filter dropdowns for each tab of a report
  • Export of wordclouds as image

Reports and Dashboards

  • Placeholder element in report/dashboard title
  • Read-only mode for read-only access profiles
  • Selected global filter remains across tabs
  • Automatic selection of comparison operators depending on the selected report filter
  • Individual dropdown-menus per dashboard

Settings

  • Adjustable precision of average values
  • Back-end definition of missing values
  • Set alignment and spacing in chart settings

Export features

  • Back-end configuration of pptx-exports and masters
  • Export of an entire report with multiple dashboards and charts as PPTX
  • Export of an entire report with multiple dashboards and charts as Excel
  • Export of an excel report in a pre-defined format

Metadata and import

  • Add of text fields via codebook
  • Automated incremental data update (e.g. backfill of last month, week)
  • Export of the codebook including special fields and text elements
  • Back-end function to delete and replace entire codebook

Various bug fixes and optimizations (load time, exports, imports)

Get started now and try our new version 1.9! We are looking forward to your feedback!

Rooaaarrr
DataLion

DataLion at the Digital Challenge 2018

In mid-June, numerous specialists and interested parties from agencies, institutes and companies got together in Munich in order to discuss the latest developments in digitization.

We were very much excited to be there as well: our CEO and founder Dr. Benedikt Köhler introduced the new world of data science. In a live example he demonstrated how one can use Foursquare to avoid being marked as a tourist when visiting an unknown new city.

The speeches and presentations where up-to-date and rather widespread thematically. The participants were presented with examples of the applications of AI and Blockchain, as well as Best and Worst Cases of Influencer Marketing.

Johannes Ceh´s keynote topped off the conference by highlighting the opportunities and risks of digitization. The networking afterwards lived up to the expectations and the speeches were discussed thoroughly at the After Show Party.

The Five Most Creative Music Visualizations

“Data is power”: by now an established fact. “Music is power”: a universal truth. So what do we get, when we creatively combine the plethora of available data on music with analyzing techniques and powerful software? We gain knowledge, insight and tremendous inspiration. We have gathered some really interesting and inspiring interactive and non-interactive visualizations on the evolution of music as well as on current distinctive trends. Enjoy!

Using data from the Billboard Top 100 Data 1958 – 2016, The Pudding´s Matt Daniels  has visualized the evolution of music taste on a month-to-month basis over the years: every top 5 song from 1958 to 2016 in the U.S.. Headphones required – it´s a “soundalized” visualization!

Kaylin Pavlik, on the on the other hand, in her Blog post „50 Years of Pop Music” has used R and data from the Billboard Year-End Hot 100 to offer a more quantitative insight into the evolution of Pop from 1965 to 2015 in the U.S..

Colin Morris (again of The Pudding) is putting one hypothesis into question: Are Pop Lyrics Getting More Repetitive? . The author used the Lempel-Ziv algorithm to measure repetitive lyrics with compression. The higher the compressibility of a song, the more repetitive its lyrics are. So seems like Rihanna´s Lyrics aren’t as much of a poetry as Frank Sinatra´s used to be! Wait, we did see that coming, didn´t we?

On a more international level, Spotify and open source platform CARTO have created the “Musical Map of the World”. The interactive map is making use of Spotify´s data from cities all over the world to give them their special “musical character”. You want to know how Germany sounds like? Click on the country to listen to its distinctive music!

Brady Fowler of Decibels and Decimals is not a friend of classifying music through genres. So he used Spotify data and Python iGraph to visualize the connection between artists and their music. The beautiful graph is the outcome of clustering the artists into groups, based on the listening habits of Spotify users.

Did you find the visualizations as interesting as we did? What would you add on the list? We are looking forward to your input.

DataLion out and about in Munich

We are proud to be holding speeches in two events in just one week. It all starts on Monday, the 11th of June at the Travel Industry Club. Our CEO and founder Dr. Benedikt Köhler will be talking about how the travel industry can make use of smart data. The event is taking place at Schweiger´s Kochschule and starts at 06.30 pm. You can find more information here.

AI, AR/VR, Blockchain, Influencer Marketing, GDPR – all these current trends will be discussed during the “Digital Challenge 2018” on Thursday, the 14 th of June. DataLion is right in the middle of it, with Dr. Benedikt Köhler talking about Alien Data Science (no worries, E.T. will not be showing up ;)). The location is the Freiheizhalle, starting at 9 am. Read more about it here.

Best 4 free math books for deepening your machine learning skills

The best things in life aren’t things – but free books. At least if you want to spend the next few weeks to take your machine learning to another level.
We’ve selected five great books that help you to understand one important aspect in machine learning in a very profound way. Thanks to the Open Access initiative, all of these works are available for free:

Elements of Statistical Learning

The Elements of Statistical Learning by Trevor Hastie, Robert Tibshirani and Jerome Friedman

If you need a refresher on your machine learning methods, ESL is the book to go to. From Lasso to boosted trees and ensemble learning – this beautiful typeset work covers all the bases for your professional life as a data scientist.

Here’s the latest 12th printing from January 2017 as PDF download.

Convex Optimization by Stephen Boyd and Lieven Vandenberghe

Convex Optimization

If you have spend some years in machine learning, the probability is very high, that you’ve stumbled upon convex optimization problems. The theory and methods around convex optimization has been around a long time. But until a few decades, they were thought to be mostly of theoretical value. Today, convex optimization is e.g. an important part of Deep Learning and many smart things around are powered by these algorithms.

Here you can download the full book by the Stanford professor for free – and there’s a lot of additional material on the website and even an online course.

Group Representations in Probability and Statistics by Persi Diaconis

Group Representations in Probability and Statistics

This book goes back to Diaconis’ lecture notes for his course on this topic at Harvard in the 1980s. There are a lot of situations where data scientists have to deal with rankings, e.g. consumers having to rank products in a survey. These mathematical problems can be solved by applying group theory.

But wait, there’s more: As Diaconis is also a magician, the shuffling of cards also plays quite a role in this work.

The book is available at project euclid.

Mining of Massive Datasets by Jure Leskovec, Anand Rajaraman and Jeff Ullman

Mining of Massive Datasets

And finally, for something more accessible: This book is already a classic on big data techniques for processing large data sets. If you haven’t already, you really take a look at the 2nd edition of the book that also includes mining large graphs and map-reduce programming. You can also take a look at the first chapters of the upcoming 3rd edition.

Here’s the book’s homepage with a lot of additional information.

I hope, there are some useful additions to your reading list. Looking forward to your feedback about the books. What was especially useful? Which books are missing from this list?

Craft Beer & Data Tasting

After the great success of our first “Festbier Tasting” side-event of the Bits & Pretzels startup conference, we are looking forward to host another data-driven beer tasting to celebrate our first trade-fair appearance at the world’s largest marketing research fair Research & Results.

This time, we will explore some of the most fascinating Bavarian craft beers. Together with beer sommelier Stefan Hermansdorfer, we will taste six very different craft beer varieties from Munich and region.

Of course, all our tasting results – from appearance and mouth-feel to aroma and taste – will be visualized in real-time on a DataLion live dashboard:

We are looking forward to this event – and there are still a few remaining free tickets available: Register for free at Eventbrite.

When: Wednesday, October 25, 2017
Time: 8pm – 10pm
Where: DataLion GmbH, Herzog-Wilhelm-Straße 1, 80331 Munich (at Karlsplatz/Stachus)

Data-Driven Beer Research @ DataLion

Monday evening was a special evening at the DataLion office in Munich:

We hosted a beer tasting event as an official side-event of the Bits & Pretzels Founders Conference and Startup Night.

Together with beer sommelier Stefan Hermansdorfer, we tasted six different beers from Munich and surroundings that were all in the Märzen or Festbier style (Hacker-Pschorr, Augustiner, Hofbräu, Tilmans Das Helle, Eittinger and Giesinger). One of them even had a lion on the label. And because we all love data, we also created a short survey for describing and scoring the beers in terms of appearance, aroma and, of course, taste.

All results were displayed in a DataLion real-time dashboard in our conferencing room: We chose radar charts for our live visualization because they allow to compare the different taste profiles of the beers very intuitively at one glance. Every time one of our guests rated a beer, the results were updating and showing the new rankings.

You can see the live dashboard at this link. If you’re interested in live dashboards that are connected to a quick survey for your business, just leave a message at our contact page.

Big Data Design Thinking

Solving problems and developing innovations for Big Data with Design Thinking

Data is the new oil. But while the use of oil was relatively clearly defined, there are much more possibilities in data. It’s an endless story.

Almost all companies are facing the challenge of having tons of data from very heterogeneous sources. But often, they lack a clear vision of what the data could be used for.

What are new business models that can be fueled with the data? Which data products can be defined and sold by the company? The solution is: Design Thinking.

Continue reading “Big Data Design Thinking”