Stacked bar charts? Mixed feelings!

Stacked Bar Charts

Part-to-whole analysis is a common task in business. Let’s say we want to analyze how much different product groups contribute towards total revenue. Or we want to analyze our cost across different cost element groups. One way to do this visually is to leverage waterfall or pareto charts. Another popular option is to use stacked bar charts. Stacked bar charts are just a special type of bar chart. Instead of spreading the different categories out across the x or y axis, we stack them. But are they really useful? I have mixed feelings about them.

Single Stacked Bar Charts

Below is an example of a stacked bar chart. This provides an overview of the cost structure for a certain fiscal quarter. You can see that each stack in the chart represents a specific cost element group. The entire stack indicates the total cost.

Is this a good chart? Sort of. Notice how much effort is involved in reading the graph. Comparing the individual stacks requires effort (look at Commissions and Travel, for example). I also find it hard to read the specific values of each stack (how much was spent on advertising?).  On the positive side, this chart allows me to quickly identify the total cost. And I obtain a somewhat solid overview of how the money was spent. However, most business analytics platforms like Cognos 10 allow you to hover over a chart section to see the individual values. That makes the stacked bar chart above somewhat more useful.

Another way to display this – and I prefer this option – is to use a regular bar chart. Take a look:

Bar Chart CognosNotice that the comparison of the different cost categories is a lot easier. You can quickly read the individual values and the comparison between the cost elements is easy as well. But I am missing the total of my cost. We would either have to calculate that or include the information in a different manner. The stacked bar chart therefore does not really impress me in this type of setting.

Multiple Instances

So, should we toss those stacked bar charts then? Not necessarily. Take a look the next example. The analysis is now extended to the entire fiscal year. There are multiple instances of the stacked bar chart.

Stacked Bar Charts Cognos

Notice how different this looks. You can quickly see that total cost have increased in the last quarter (after decreasing slightly). I am also able to see how the different cost elements have changed throughout the fiscal year (look at advertising, for example). The stacked bar chart is now much more useful. I personally like this. But what about the traditional bar chart in this situation? Let’s take a look:

Multiple Bar Charts

The graph invites you to compare the cost composition quarter by quarter. The comparison between different quarters is also not difficult. The only problem with this version is that the overall cost are difficult to assess.  Both versions have their strengths and weaknesses.

Stacked Bar Charts – Summary

Stacked bar charts are certainly not bad. But as the examples above show, they are stronger in a multi-instance setting. But even then, you need to be careful: stacked bar graphs tend to look strange when you have negative values (give it a try!). The single stack is not that strong as compared to the traditional bar chart. Both offer different insights. And let’s not forget about waterfall and pareto charts as well.

From an analysis point of view, I would probably want to switch between the different charts. IBM Cognos 10 provide users with the ability to change chart types on the fly. That makes the analysis of data very interactive.

Have you added stacked bar charts to your toolbox? If yes, make sure to use them in the right circumstances.

P.S.: I will take a look at stacked area charts in February.


Guest post about data visualization on

If you follow this blog you might remember that Mike Duncan from the small business consulting firm Bizzeness posted two guest entries back in August. Mike shared some interesting views about Dashboarding and the selection of proper KPIs. Those were amongst the most popular entries this summer.


Today I was finally able to return the favor. Check out the guest post on the Bizzeness blog. The article is about one of my favorite topics: Data Visualization. While you are on that site, make sure to take a good look at their blog. There is great content!


Why you need Many Eyes


Have you heard of Many Eyes? Strange name, huh? Well, Many Eyes is a pretty cool and simple service on the web that let’s you visualize and explore data. Need a quick word cloud of Steve Job’s last speech? Sure. Just go to Many Eyes and create it. Need a quick scatter plot of your hear rate and running pace? Just go to Many Eyes and create it there.

IBM Many Eyes

Dashboarding Sticky Visualization

A few thoughts about gauge charts

Let me start by saying an obvious thing: Gauge charts do deserve some recognition. Actually, they deserve a medal – a medal for being the most controversial chart type in history.

When I first saw a gauge chart, I was impressed. They looked pretty cool. Things changed when I got involved with my first BI project. A closer look revealed that these charts are actually pretty tough to look at. But let’s back up and start from the beginning.


Gauge charts allow us to visualize data in a way that resembles a real-life speedometer needle or a regular gauge. They usually  display a single key measure. The outer scale of the gauge is often color-coded to provide additional performance context (green for good, red for bad). Below is a typical example:

Gauge Charts
Feels like driving a car?

Data + Powerpoint = Wasted Time?

Presenting data can confuse people

Let me be blunt and honest: Too many presentations and their accompanying slide decks absolutely suck. And they especially suck when it comes to displaying and discussing data. Over the past few years, I have sat through days- worth of boring and utterly useless presentations. Such a waste! And there was so much potential: great data points and valuable information. But all this was well hidden behind complex and confusing charts. And believe it or not: that is a problem for business analytics.


We spend so much time and money on implementing business analytics software. We create so many awesome reports and dashboards. There is so much potential. But way too many people take this valuable information and literally destroy it by using the trusted information to create useless and complex slides. Those slides are then presented in meetings where we try to sell ideas and we where try to make collective decisions.  But due to the convoluted slides (often coupled with poor communication skills) most messages fall flat on their face. I am tempted to say that the ‘last mile’ of business analytics is broken in these cases. It’s about time to fix that.


Dense & confusing - A typical slide?

Famous statistician and popular data guru Hans Rosling famously discussed this issue and compared the presentation of data to playing music: “…few people will appreciate the music if I just show them the notes. Most of us need to listen to the music to understand how beautiful it is. But often that’s how we present statistics: we just show the notes, we don’t play the music.” It is not enough to create a sophistcated data warehouse and some shiny reports. No, we need to make the data sing when we present it to other people an especially larger audiences. Developing solid presentation skills should therefore be high up on the priority list for anybody who works in the area of business analytics.


There is a big difference between presenting insights to an audience (meetings, events, etc.) and analyzing data at our desks. Following a presentation requires a different level of energy and focus: it is a lot harder to follow in many cases. Our brain tries to juggle processing the information on the slides while listening to the speaker. We therefore need to make it easy for our audiences to receive the messages that we have found and prepared. The reports and charts that work at our desks do not necessarily work in a meeting room. We have to think differently. And that’s the disconnect we often see and that Hans Rosling aludes to: we do not think differently and simply show confusing details when we should be showing a clear story. We are short-selling our efforts and the impact of our insights in effect.

But there is good news. Learning how to present and how to tell an inspiring story using data in a presentation does not have to be complicated. In a few days from now I will share some tips & tricks that you can put to immediate use. Start thinking about those presentations! As always, I am curious to find out what your experiences with this are.