Waterfall charts, pareto analysis and beyond

Waterfall Charts?

Do you like pies? I do! But not for analyzing data. Pie charts are just too busy and too hard to read in most situations. Yet they are the frequent tool of choice for visualizing the composition of a certain variable. Take a look at this example: I want to find out how total revenue is split between different product categories. The easy choice would be to create the following chart:

A product pie - created with IBM Many Eyes

Personally, I don’t like this! You constantly jump around the different pieces and it is really tough to obtain the overview that we are looking for. Pie charts work for few variables that account for larger pie slices. Another and better option is to use a bar chart. To make it easy for the viewer, I ranked the values in IBM Cognos 10:

Ranked bar chart - much better!

Isn’t this much better? You can quickly identify the best performing products. It is easy to read the individual contribution of each product. But I am not able to see my total revenue amount. Good news – there are different ways to perform this type of part-to-whole analysis

CHASING WATERFALLS

Waterfall charts (sometimes called Progressive Charts) allow us to visualize the composition of the a value along different segments. In plain English: Total revenue can be visualized as a build up of the different individual values of the products. Take a look at the example below. The values are once again ranked:

Waterfall Chart
A classic waterfall chart: Total revenue across products

Notice how quickly you can see the segment revenue along with the total revenue of roughly 540 million. I personally like the clean and uncluttered look. Reading the individual values is a bit harder than in the traditional bar chart. But IBM Cognos 10 allows me to hover over each segment to obtain the actual value. It’s personal taste which chart is more effective. Here you can see both versions side-by-side.

Which chart is your favorite?

Waterfall charts are great for any kind of part-to-whole analysis: revenue/ margin across products, customers, channels.  Composition of Profit across the P&L, etc.. There are a few interesting examples on Wikipedia.

THE PARETO CHART

There is yet another great option for displaying this type of data: Pareto Charts. They are are named after the Vilfredo Pareto who proposed the 80/20 principle. This chart simply enriches the bar chart we saw earlier with a cumulative percentage line. Take a look:

Pareto Chart
The Pareto chart: The cumulative percentage line adds further context

This graph allows us to quickly answer questions such as: Which products create 80% of my total revenue? (roughly everything up to Alpha Bronze). In other words: the pareto chart allows me to identify the most important items.  IBM Cognos 10 allows different customization options which I would highly recommend looking at. Hiding one of the axis is not a bad idea, for example. The use of colors could be helpful here to identify the main product categories:

Colors allow the user to identify the main product categories

Waterfall Chart vs Pareto Graph?

Seems like a lot of options, right? Which one is the best? Can’t say. It really depends on the situation. I like all of them. There are slight differences and it depends on the user. Waterfall charts are probably better suited for executive dashboards. Pareto charts are more likely useful for analytical purposes. And the classic bar chart is always is a winner. Try to add these charts to your tool-box! And drop those pies.

Scatter charts – Good for relationships

SCATTER CHARTS – THE BASIC IDEA

One of the interesting and really fun things to watch is young kids learning about cause and effect. They pull on a string and music starts playing. They giggle. They push a car and the car begins to roll. And so we learn to be curious at an early age and we learn to look for cause and effect relationships. And this an especially useful skill to have in business. The discovery of relationships can help us make better decisions: what happens to our revenue, if we increase marketing spending, what happens to our customer inquiries if we lowered the price of our top product? Answers to these questions can provide valuable insights.

WHAT IS THE RELATIONSHIP?

How can we go about testing and identifying these relationships? One option would be to combine two data sets in a chart. Let’s say we wanted to analyze the relationship between our price and customer inquiries. How do customers react to a price increase or decrease? We could create a combination chart for our products which outlines the price (red line) and the inquiries (green bars):

Product Alpha - Is there a correlation between price & inquiries?
The product Charger - higher correlation?

When you look at these charts it seems that there is a loose relationship between price and inquiries for Alpha but a stronger relationship for the Charger product. But it is hard to really tell. Especially for Alpha. Overall, this chart is not all that useful. We need more information.

SCATTER CHARTS

This is where scatter charts come in handy – they allow us to quickly analyze the relationship between two numeric variables. We basically take a regular Cartesian 2D coordinate system with our two numeric variables plotted on each axis. The general norm is to plot the independent variable (in this example the price) on the horizontal axis and the dependent variable on the vertical axis (customer inquiries). Here is a simple example that I created with IBM Many Eyes:

Scatter charts
Marketing Spending and Revenue

The dots represent the values for individual marketing campaigns. We mark the amount of  spending on the x-axis and the resulting revenue from the campaign on the y-axis. We can easily tell that there is a relationship between marketing spending and revenue – we could almost draw a line between the dots. There is just one outlier on the bottom of the right-hand side.  By the way, it is really easy to create these charts with IBM’s Many Eyes tool. Check it out when you get a chance!

PRICE AND INQUIRIES

Back to the initial problem. Let’s see how price sensitive Alpha and Charger are. Let’s take a look at the resulting scatter plots. We have created these charts in Cognos 10 using the same data set. We have also included a trend line to make it easier to see a potential correlation:

Scatter charts in Cognos 10
Same data in a scatter chart. This makes more sense

Both of these graphs now tell a clear story: Alpha’s dots are literally scattered throughout the chart. There are plenty of outliers. This shows that there is just a weak correlation between price & inquiries. The picture is different for charger: The dots are more clustered and we can draw a good line, i.e. the correlation is pretty high.

SO WHAT?

Scatter charts are pretty simple to create and they do tell a good story if used for the right purpose. They are also ideal for large data volumes. However, they do ignore time. The combination charts I showed above would do a better job at that. But if we want to focus solely on the relationship, the scatter plots are better suited. Even though scatter plots are relatively easy to read, I would not recommend using them in an executive dashboard. You definitely need to know how to use them. They are probably better suited for analytical people. Also, keep in mind that while these charts help identify relationships pretty well there might still be other influence factors. But that is really common sense. So, next time you want to explore your data in a different way try scatter charts!

 

What cartoons can teach us about information delivery

A picture says more than a thousand words, right? Managers drown in pages of numeric reports. But as John Medina, author of the famous book ‘Brain Rules’ clearly pointed out: “Vision trumps all other senses“. In other words: we are much better at absorbing information through visuals than we are at reading numbers and letters.

To see how effective visuals can be at delivering complex information we just have to look at something that we are all surrounded by: Cartoons. They are in newspapers, they are on websites, they show up on twitter. Cartoons are drawings. A single picture that tells a humorous or critical story. Once you look at them carefully you will notice how deep and smart they can be: They tell a thoughtful joke or story in a single picture.

Talk about the power of visualization! Remember the scandal about the Danish cartoonist a few years ago? A single picture set a huge scandal in motion. To convey the same message, a writer would have to fill a lot of pages. And those pages wouldn’t be all that powerful.

To learn more about cartoons I ended up watching this short Ted presentation by Patrick Chappatte. It is quite entertaining and it highlights the power of visualization. We should all strive to learn something from this. Shouldn’t we all use more visuals in our daily lives? Should we toss those endless 2-dimensional reports and replace them with good, solid visuals that tell a clear story? You know my answer.

 

Why bubble charts are cool

Things can be complex. Especially when we look at multi-dimensional data-sets. The objective of charts is to visualize data in the most effective and easy way. You shouldn’t need a PH.D. degree to decipher a complex chart. But it happens. There are a lot of complicated and useless charts out in dashboards. And it happens more often than we think. For example: Once we reach more than 2 dimensions, many people reach out for 3D charts. Let’s say we want to analyze market size, market share & margin. Many people are tempted to simply create a 3D bar-chart like the one below:

Chart 1 - Is this useful?

There are a lot of obvious problems with these type of charts: The dimension have different scales and it is therefore impossible to decipher. And let’s be honest – this looks super ugly. I could not, would not make a decision based on this chart. The other option is to break this out into multiple charts. But that requires a lot of space – and space is tight in a good dashboard. Analyzing numbers would also be more difficult in that setup as we have to shift our view from one chart to the next.

THE CASE FOR BUBBLE CHARTS

There is a better way to display this type of data. My boys loves this chart type: Bubble charts (all kids love bubbles!). Bubble charts allow us to visualize three different measures at the same time. And not only that: they are easy to read and they allow us to make critical associations between these measures. Let’s have a look at an example: This is a classic bubble chart that displays three different measures: Late shipments, damaged shipments and shipping cost for different carriers. The first two measures are obvious – they are represented by the x and y axis. The shipping costs, however, are visualized via the size of the bubble.

Bubble Charts
Chart 2 - A classic bubble chart

Notice how easy it is to read this chart (which vendor has the best performance?). Depending on the problem that I am trying to solve, I could simple look at the top right area to find the black sheep that are super later and also careless. Or, I could first focus on the size of business that we do with each carrier by picking out the large bubbles. Pretty simple. Also notice how this chart allows me to combine three measures with different types scales: percentages and absolute values. The traditional 3D bar chart was useless.

THERE IS MORE

In Cognos 10, we can also turn any bubble chart into a quadrant chart. This is useful if you want to categorize your data a little further by using a common layout like it is used in a SWOT or market attractiveness analysis. Take a look at the bubble chart that we created using the data from the first Excel 3D example:

A classic bubble chart
Chart 3 - A quadrant bubble chart

This puts the data into further context and makes it really easy for managers to spot specific key points. For example, the attractive markets (high margin & high market share) are up in the right upper corner.

Cognos 10 also allows you to hover over each bubble and you will get the numeric details behind each bubble. This makes it really easy to explore the data.

Chart 4 - Use your mouse to explore

THE LIMITATIONS

As nice as the bubble charts are, they are certainly not perfect. Take a look at Chart 3 above and focus on the intersection of 11.5% Net Margin & 2% Market Share. There is a bigger bubble covering a smaller one. That can easily happen. A superficial glance over the chart can therefore be problematic because we would not notice this. Careful color choice could potentially help uncover these cases. This probably also highlights that bubble charts might not be an ideal solution for large data sets as there would be too many overlaps. But nothing is perfect, right?

cognos slider
A slider - easy to use

Also, keep in mind that bubble charts in their pure and simple form only provide a snap-shot in time. Time-series analysis has to be done in a different manner. But the good news is that Cognos 10 offers us sliders. We can use these sliders to walk through history and easily discover changes in the data.

LAST BUT NOT LEAST

One person who has really popularized the bubble charts is scientist Hans Rosling. He literally makes data fly. If you haven’t done so, make sure to watch one of his famous TED presentations.

Take a look at bubble charts! Consider them for your next project. They are easy to understand and they allow us to make critical associations. Chances are managers who have attended business school will certainly like them.  A friend of mine always says that managers are like kids. And kids like bubbles, right?

[twitter-follow screen_name=’cpapenfuss’]

Why sparklines sparkle in a dashboard

SPARKLINES SPARKLE

One of the key features of the dashboard in any car is the fact that it is on a ‘single page’. Car dashboards are usually very simple and it doesn’t take us more than a few seconds to familiarize ourselves in a new vehicle. Could you imagine driving a car with a complicated dashboard? Could you imagine having to scroll through multiple pages to get a speed reading while you are driving down the German Autobahn with 200 km/h? The answer is no.

THE CONSTRAINTS

Good management dashboards should be concise as well. Executives especially don’t have the time and patience to scroll through multiple pages of information. But that often creates a problem: We have a lot of information to show but there is limited space. This problem arises especially when we want to show trends. A typical dashboard might want to show profits, margins, win/loss ratios, revenue, pipeline across multiple regions. One way to display this rich set of data in a table. But tables are ineffective in quickly displaying trends. And they take up a lot of space. The other option would be to show multiple line charts. But those take up a lot of space as well.

THE SPARKLINE

Information design guru Edward Tufte developed a solution for this problem. He created a new but very simple chart called a ‘Sparkline‘. He describes them as “data-intense, design-simple, word-sized graphics”. In other words: sparkline charts are very small, yet they display vital trend data. Take a regular time series such as revenue by months. Stick that into a sparkline chart. Below is an example:

Simple sparkline charts

ELEMENTS

When you look at the example above, you will notice how easy it is to spot the general trend: the Conservative Party is loosing ground. You have also noticed the little dots. IBM Cognos 10 allows you to color code the low, mid and high point. (red, black, green). The Green Party is gaining. You can hover your mouse over the line to get the actual value. This type of chart simply provides the ability to obtain a quick overview of the general trend and all that within a very tiny amount of space.

SPARKLINES AT WORK

Sparklines truly show their full potential when you supplement them with metrics or other charts. Take a look the next example:

Sparklines in action

The above object allows us to review margins across regions. The sparklines quickly display the trend. The trend is supplemented with more detailed information. Notice how concise and compact the total object is. And here is an example where we can see the sparklines integrated into a regular dashboard:

A sample IBM Cognos 10 dashboard

The upper part allows us to quickly provide an overview of critical metrics across different dimensions.

SPACE SAVERS

In a prior blog post I discussed the bullet chart. Those charts are also quite small and highly useful. When you combine the sparkline with the bullet chart you have a powerful combination. Notice how much information we are able to absorb (trend, actual value, current performance). A traditional line chart would take up a lot more space.

Sparkling Bullets

SPARKLINES – A FINAL WORD

Sparklines are great for displaying trends. This is ideal for supplementing current information (e.g. YTD Sales). They are tiny, they are easy to use and they are easy to understand. However, they will not replace a traditional line chart. Line charts are definitely a better choice when you want to perform a more detailed analysis. Next time you build a dashboard try to incorporate some sparkling charts.

 

Hands-up! Here comes a bullet chart

Executives often want to have a quick overview of some key metrics to find out what the state of their business is. Questions like: Are my margins on target? Is customer satisfaction within an acceptable range? What is the size of my pipeline? All this information is typically summarized in a dashboard. The easy thing is to simply put this into a table. But tables are hard to read. And they take up a lot of space. The other popular option is to put the data into a gauge chart. But gauge charts take up a lot of space, too. And if we are honest with ourselves: they are really hard to read. Granted they look cool. But do they tell a story in an effective and efficient manner? Probably not.

A BULLET

A few years ago, Stephen Few introduced a new chart that promises to fix the shortcomings of the above described approaches. It is very simple but powerful chart and it is called ‘Bullet Chart’. Cognos 10 allows us to leverage these charts. Below is an example:

Let’s take a look. The above example shows the election results of a fictitious political party. The blue bar in the middle indicates the actual value (4.8%). The short black bullet towards the top indicates a target measure (e.g. budget, forecast, etc.). Color shades display ranges of performance (e.g. poor, acceptable, good). We can quickly see that the party missed the target but the result falls into the acceptable range. It is indeed a very simple chart that provides a lot of information in a concise manner: Target, Actual, Performance Rating. In Cognos 10, you have the ability to set five different performance zones in different shades or colors.

A STACK OF BULLETS

One the things I like about the bullet charts is the fact that you can easily stack them. That makes the bullet chart an ideal way of communicating multiple measures in a dashboard. It is easy to get a quick overview and the stack is very space efficient. In Cognos 10, we can have vertical or horizontal bullet charts.

The stack of bullet consumes little space

CAUTION?

The are just two minor downsides that I see with the bullet charts. The chart in its pure form does not allow us to show the future trend or forecast. Also,  it does not display history. It is a simple snapshot in time. But the last shortcoming can easily be mitigated by combining the bullet chart with a sparkline. (I will look at sparklines in the next post)

Toss those gauges! Take a look at the bullet charts next time you design a dashboard. But make sure to train your users. Despite its simplicity, I have seen some people struggle to understand this chart. We don’t want our users to bite another bullet, right?

Charts? Yes, I know….

Dashboarding….Part 2.

Yes, we do know that we shouldn’t eat those chips. Yes, we do know that we should read more books. Yes, we know that we shouldn’t drink that much coffee. But…..BUT…..We all know that there are small things we can do that could really have a big impact on our life or job. But we still don’t do these extra little things but they take time and effort. We are all guilty of that. I am for sure. Most of us can tell some stories about that.

CRITICAL SKILLS

About a year ago I had a revelation. Together with two hundred other business and IT people, I was listening to a presentation about common data visualization mistakes. YAWN…. How exciting. A few minutes into the presentation it dawned on me. I actually didn’t know all that much about proper charting techniques. I had never actually paid much attention to it. By looking at the reaction of the other attendees around me, I noticed that they were in the same boat. How can that be? We are all professionals that are dealing with data on a daily basis. Yet, so many of us have never paid much attention to proper visualization techniques.

THE MAGIC PIE

What is the message?

That day I decided to make a change.  I felt a pressing need to learn more about charting techniques. And when you look around you can easily see that many of us need to make this change as well: We use ugly pie charts left and right, we create meaningless 3D visualizations and we connect data points that should not be connected. And that is not too surprising. Nobody ever really taught us how to properly visualize data and when to leverage which type of chart. To make things worse, nobody ever questions this. Executives seem happy with their colorful 3D charts. I will never forget the day when my then-boss at a traditional German company almost fell off his chair when saw a ‘cool’ 3D diagram I had produced with Lotus 1-2-3. But at the end of the day, many charts do a very poor job at telling a good story about the data.

DATA IS THE NEW OIL

Data volumes grow. The speed and volatility of business are increasing. As a result, we all need to make sure that we find meaningful trends and insights in our growing pool of data. There is a lot of insight to be unlocked. Many people are therefore saying that ‘data is the new oil’. But in order to really get the best out of our data we need to learn how to visualize it properly.

THE POWER OF VISUALIZATION

As John Medina points out in his bestselling book ‘Brain Rules’: “Vision trumps all other senses.” The nerve pathways of our eyes to brain are extremely powerful. John Medina continuous by saying that “Professionals everywhere need to know about the incredible inefficiency of text-based information and the incredible effects of images”. In other words: dry tables of numbers just don’t cut it. A picture says more than a thousand words. Let’s create charts that tell a powerful story. And those charts are especially useful when we utilize them in dashboards. Dashboards should be visual to allow users to quickly absorb and digest critical information.

CHANGE

Charts can convey a powerful message

We should therefore all take the time to learn more about charting techniques. And it’s a simple thing to do. Pick up a copy of one of Stephen Few’s books (see below for a list of recommended reading). Roam around the website of visualization artist David McCandless (I wrote about him a while ago). Play with your current numeric reports and put them into charts. Compare the different stories. Over the next few weeks, I will write some posts about some of the powerful charting options in IBM Cognos 10. IBM Cognos 10 comes with almost 160 different chart types. There are some fantastic tools in there that can really make your data fly. Make sure to follow along to get some tips and tricks.

Recommended reading:

What photography taught me about dashboards


Photography and dashboards? Huh? Fire and Ice?

Photography is a big and important hobby of mine. And it is a tough hobby. There is a lot to learn and the opportunity to make mistakes (read: create photographs that really suck) is huge. It starts with understanding your camera, deciphering basics like Aperture, Shutter Speed and ISO. But the hardest thing for me is photographic composition. Composition focuses on how we design a photograph. Over the past few years, I have studied many master photographers and read a bunch of books trying to educate myself and to improve my pictures (a tough mission as my artistic brain is completely underdeveloped). A few weeks ago, I realized that photographic composition can teach us a few things about Dashboard design. Dashboards should be highly visual after all and they need to convey information in a short-period of time.

  1. Less is more: Many successful pictures have been reduced to a bare minimum. Each element in the frame has a distinct purpose. You will hardly ever find a great photograph which contains empty coke bottles lying around for no purpose. It would create a distraction. By reducing the elements in the frame a photographer creates focus. The same is true for dashboards. We have so much information available. People therefore try to cram as much into a dashboard as they can. And we stick logos, banners etc in there along with messy charts and reports. But less is more. If we reduce the building blocks to a minimum, we can help managers focus on the important things.
  2. Arrange carefully: Successful photographs are able to convey a certain message. The message is crafted by arranging the elements in the frame in a certain way. In other words: we can’t just find a nice scene or object but we need to carefully consider where to place items. The same applies to dashboards. Stephen Few for example points out that we should place the most important block in the upper left corner. That’s where the Western world starts reading. That way we can ensure that managers focus on the most important element first. Also, we can employ different techniques to direct our eyes. (see some examples below).
  3. Choose colors wisely: Different colors communicate different things. Our eyes focus on bright elements before they refocus on darker elements for example. Red or bold elements alert the eye as opposed to darker colors or thinner elements. A great photograph therefore utilizes colors with purpose. Sometimes colors take away from the meaning of a photograph. Black & White would be the obvious choice in that case. When it comes to dashboards we should employ the same considerations. Colors and fonts should be used with careful consideration. Too many dashboards are colorful without a specific purpose and it confuses the message. “What should I look at? I can’t see the tree in the forest.” Careful color choice helps direct the attention to the important items. For example, you can highlight an exception in red. But color preference is also a personal choice.

Here is an example where we can see these principles at work:

The overall scene

The scene is quite busy. The yellow color is not really useful. Not a good photograph. Basically a typical snapshot.

This photo creates focus. There is only one element in the frame. Much better. But it is kind of boring.

This photo is much better. It is more dynamic. Same object, different placement. What a difference!

Yet another version. This photo simply works. It is dynamic and the object is placed in the right spot. Color is not needed in this case. The picture works as a black & white.

Here are some examples that highlight how these things apply to dashboard design. Consider each box as a representation of an object (chart, query, etc.). Look at the first really sloppy design attempt. Looks weird? Well, it does happen quite a bit. Take a look at this beautiful collection of crazy, busy dashboards collected and displayed by Hichert & Partner.

Performance Dashboard
Want some candy???

And now look at the following layouts below. They are simple. There is no added noise that distracts. Notice that the light blue matches the design of this blog……Notice how the careful placement of the boxes makes a difference. The lines indicate how a typical user walks through the content. Also, note how the use of color changes things.

Dashboard DesignNext time you develop a dashboard be careful with your design. It does make a huge difference. Just these three things alone can have a big impact on the effectiveness of your dashboard.

If you are interested in this topic, please get in touch with me. We will be running some workshops about this topic across Europe in Q1 & Q2. Also, I will create a few additional posts about enhancing your dashboards with great charts over the next few weeks. Make sure to come back here!

“What you communicate depends both on the blocks you select and on the ways you arrange them”,

Freeman Patterson, Master Photographer

 

Visualization of unstructured information

The other day, I attended a great exhibit that was put together by my friend Bernhard Lermann (@lermann) and his companion Jens Semjan. Their Munich exhibit called “The Eternal Timeline Show” is highly unusual, very entertaining and informative. Apart from educating me on a few things, the exhibit certainly triggered a lot of thinking around data visualization.

THE ETERNAL TIMELINE

Eternal Timelines

What is so special about this exhibit and what is the relationship to visualization? Very simple. Lermann & Semjan decided to visualize history. We have all read history books. Many of them are quite boring and lengthy. The two artists decided to take a different approach by creating visual timelines around certain topics. In the exhibit you can find a timeline for Apple, for example. Another timeline focuses on the history of networking. All this is displayed on long sheets of endless paper. The spectators are able to walk the timelines up and down. Apart from some classic history themes (epidemics, revolution, etc..) there are also some interesting newer topics such as tweets that led up to the revolution Egypt a few weeks ago.

ENGAGING VISUALS

The timelines represent an excellent example for visualizing unstructured information. What I found is that many people walked the timelines up and down. And many people showed a reaction: They stopped, they thought, they discussed. It was quite interesting to watch. But not surprising. We all know that pictures say more than words. Being able to physically walk the timelines spectators are getting engaged with history. They are also able to relate events (just move over to the left…). Last but not least, I found that I had absorbed a lot of knowledge within a very short period of time. Visuals do that for us. They help us make connections.

THE LESSON?

You Tube was here! The history of networking

Visuals are extremely helpful. They help us identify trends, see patterns and they sometimes connect with us on an emotional level. There is plenty of room for us to leverage this powerful tool in business. Let’s start visualizing our monthly variance reports. Let’s visualize the behavior of our customers. We can all save a lot of time and we can have more fun at work. There is ample of opportunity for us to visualize our data and information. This exhibit is a fine example of that. It inspired me and many other people. If you happen to be in Munich, make sure to stop by the exhibit or get in touch with the artists. Also, make sure to watch David McCandless’ presentation about data visualization at TED.

The power of Data visualization

We all know that visual information display is more effective. But yet, we still create, run & analyze hundreds of thousands of flat reports. A survey from 2007 (CFO Executive Board) found that the average management report is 30-40 pages long and that it contains 12000 to 15000 data points. But only 5% of those are deemed to be important. The most obvious question that comes to mind is: How much time is wasted to go through this amount of data and turn this into real insights? Would it help to start using more powerful visuals to display the hidden meaning behind all those data?

Watch this inspiring and entertaining speech by author, journalist & visualization pioneer David McCandless. I will blog about some of his ideas and how they relate to business at a later point.