Skip to Content

Visualization: Efficient Dashboard Design – Part II

/wp-content/uploads/2012/06/american_watchamacallit_114938.pngIn my last post I announced that I would analyze the – how shall I call it – maybe “the worst chart in the world”. Here it is: the origination process from a visual loser to a meaningful visualization.


I think in the first part of my series about efficient Dasboard Design, I was able to demonstrate that the graphical presentation of information is preferable to the tabular presentation. When it comes to recognizing patterns, trends or outliers charts remain unbeaten. Although tables have their place in reporting – more on this in another post!

Now the point is to show that is not sufficient for efficient Dashboard-Design to position some pie-, column charts or speedometers/gauges haphazardly on a screen page. A selection of how one should not do it can be found here.

The selection of the right chart type and the design of the chart are decisive for the expressiveness of a dashboard. I want to show this by means of an example chart in a step-by-step evaluation. But first of all I need to confess that I found the chart two years ago in a blog entry by Stephen Few with the title: „Oracle – Have you no shame?“

1_ORACLE_ORIG_en.png

In my opinion it is still probably the worst chart in the world! It seems unbelievable but this chart was – is? – the demonstrative example for the online learning course: “Oracle BI Enterprise Edition – Build Good Dashboards”! Putting it a bit bluntly, Stephen Few sums it up:


“Based on this graph, I’m guessing that Oracle now outsources the development of its courses to the primate house of the local zoo.”

Why? Beautiful 3D- cylinder charts and -lines, which make an interpretation of the underlying data nearly impossible. First of all we will delete this straining after effect along with superfluous frames. You can see the result here, built with standard components of SAP BO Dashboards
(Xcelsius). I needed to perform some sleights of hand to succeed in building the chart of step 1 with Xcelsius:

2_Xcelsius_en.pngWithout the 3D-effects and value axis starting on the same level with “0”, the basic prerequisites for a sensible interpretation of this chart are given. On closer inspection it is obvious that the “Forecasted Dollars” are not shown on the left value axes, together with the other two given values, but on the right value axis, together with the “Forecasted Units”. They can not be serious! All values („Dollars”) are depicted with identical elements – columns – and on the same axes:

3_CHECK_en.pngUp to this point I am tolerant as these mistakes are BI-standard. But it gets a lot worse: Who is the donkey responsible for stacking three annual values in order to make them visually comparable?! If I want to compare actual turnover data with data from the last year or with the forecast, I show them grouped side-by-side:

4_CHECK_en.pngFurthermore, I am able to depict the “Forecasted Units” on a uniform scale with the Dollar-values. I can do without auxiliary lines. I rather prefer the data labels on the elements. I also adopted a rule: I always show development over time on a horizontal axis from left to right,
structural development on a vertical axis. For the comparison of regional structures I therefore prefer a grouped bar- instead of a column chart.

5_SIMPLIFY_en.pngStill rather colorful. The choice of color and color-effects are deviating. Visualization gurus call this “Noise”. My advice is to use colors only if they make sense: to emphasize, as a warning signal or if the color is used to encode information. I know companies which assign certain colors to each business segment! Red and Green are reserved for “good” and “bad”. Furthermore, charts in shades of gray are easier to print. Another form of “noise” are the 6-digits chart labels: I use three digits as a maximum and scale the chart accordingly:

6_SIMPLIFY_en.pngRegular readers of my lines know: I am a fan of the regulatory framework SUCCESS by Rolf Hichert. We have these rules fully implemented in our graphomate Addon for SAP BO Dashboards (Xselsius). After applying the CHECK and SIMPLIFY rules for the scaling and design of the chart, the next step is to apply the CONDENSE rules: The goal is to reach a high density of information in order to depict complex data at a glance. The need to browse through (web) pages hinders the comprehension. On the third page, at the latest, we have forgotten what was shown on the first pages. So: minimize the chart size and put the “Forecasted Units” again in the chart. Shown in a needle chart as to set them apart from the values (“Dollars”).

7_CONDENSE_en.pngAnd finally: What is a report without a message? The name of the original chart “Regional Revenue: Current Versus Prior Period” is wrong and meaningless: I can see that it is a comparison of the turnover! Even if I do so at a second glance, as I can also see forecasted units and values.
Each report should have a message, which should be emphasized in the chart:

8_SAY_en.pngI am aware that realizing these rules of the SUCCESS-concept – SAY – is rather difficult.


To sum things up, each step of the process shown as a “slideshow” – realized with SAP BO Dashboards (Xcelsius):

The corresponding xlf-File for SAP BO Dashboards (Xselsius) can be downloaded here: The file contains the graphomate visualizations as images, so you don’t need our graphomate Addon. You may download a functional restricted Free-Version of our Addon for SAP BO Dashboards (Xcelsius) here.


Stephen Few rightly remarks, that:


“It is as if the person who created this “Good Dashboard” example of a graph did everything possible to make it as ineffective as possible.”

Terrible, really terrible, if this is sold as the “best practice” of a BI vendor!


In Part III of my series on efficient Dashboard Design I will talk about the remaining regulatory groups of the SUCCESS concept. Especially UNIFY – the idea of a standardized company-wide notation for business reports – will be a topic.

But for now I am going to eat some watchamacallits …

Pleasant holidays, Lars

3 Comments
You must be Logged on to comment or reply to a post.
  • Hi Lars,

    I really like reading your blogs but I’ve some mixed feelings about this one though. So many rules to apply!  I do not necessarily disagree with them but it’s all about context.

    Each report should have a message? We’re talking about dashboards, no static graphs with a never changing message. The graph (data) itself should convey the message.

    Why minimize the chart size in this case? Even if you were going to use this one together with many more charts, it’s still need to be legible. This one barely is. Just make them as big or small as needed. I know from experience that putting too much information a dashboard can be intimidating for certain users and thus preventing them from actually using them.

    Structural development always on a vertical axis? It can be very beneficial when using long labels (makes them easier to read) or for rankings but on the other hand they are a little bit harder to interpret since we (humans) are better in comparing lengths than widths. Also, most business users are accustomed to reading vertical oriented bar charts. 

    No colors? I know this one always sparks some controversy but again it’s about context. I very much agree with you that you should prevent noise as much a possible (Xcelsius loves noisy themes ;). However, the careful use of color can make a dashboard less intimidating and more engaging. It’s all about using color consistent and in an educated manner that fits your audience and usage scenario best.

    BTW, we’re planning to do some research on the color topic (together with a usability firm and the use of eye tracking equipment) to see how the use of color (among others) affects the user and the effectiveness of a dashboard.  

    Looking forward to your next blog!

    Frodo

    • Hi Frodo,

      thanks for your in-deep feedback!

      Let me point out some of your comments:

      > So many rules to apply!

      There will be even more rules ... 🙂 ... all for one goal: easy to read dashboards!

      One says: "A picture is worth a thousand words!"

      Most dashboards I saw need thousands words to explain, what they want to show us!

      > Each report should have a message?

      As I wrote presenting a message in dashboards is hard but not impossible - i.e. for C-Level reporting. But your are right: I should better speak of comments here. Commenting seems to be one of the most wanted functions in dashboarding.

      > Why minimize the chart size in this case?

      Because "It's all about comparison!" - as Edward Tufte states! Decision-takers need their relevant information at a glance!

      You can't compare data when preseting one chart after the other! So I minimized the chart in this post, to have more space for more charts! To be honest I think it's not small enough 🙂 But yes: this approach can be intimidating and one has to be very careful when implementing these rules - one after the other.

      Presenting before/after-examples usually convinces our clients.

      > Structural development always on a vertical axis?

      Definitely! This rule is a very easy to implement design rule  and it's facilitates comprehension of information in dashboards. I do not agree that most business users are accustomed to reading vertical oriented bar charts. There is lot's of buzz for horizontal

      bar charts to avoid pie charts!

      Time (trends) with horizontal axes, structure (comparisons) with vertical axes support the understanding of charts immediately.

      Most of our clients apply this rule companywide!

      > No colors?

      There will be colors in part III, I swear!

      You are absolutely right: Never use colors without context!

      If you are planning to do some research on the use of colors, please have a look at the work of Maureen Stone - for example here.

      Lars

      • Hi Lars,

        I really appreciate your feedback!

        I know the rules (and yes, there’re many more) and I apply them in my designs most of the times as well. I guess I just don’t like to call most of them rules since it implies that they should be obeyed all of the time and every time.

        With any design, not just dashboard design, it’s important to really understand your users and the purpose of your design. Combined with the knowledge of good design and data visualization you can create efficient, usable and engaging dashboards. But that you already know 😉

        I’m a big fan of Tufte’s work by the way. However, based on this discussion, I stumbled upon a research finding indicating that the minimalist design we are promoting isn’t always the one people prefer. I’m certainly not considering using chart junk instead but it’s a good thing to realize that what’s good in theory can be perceived differently in practice.

        That’s why I’m interested in the effects of certain design or data visualization principles on the effectiveness and perception of dashboards. My goal is not only to create the most efficient dashboard but also one that people feel comfortable using and want to use. Only then can they be really successful!

        Frodo