Perceptual Edge Discussion Forum

how to counter defense of this graph?

Fri, 10 May 2013 17:20:32 GMT

I recently received a copy of a customized report that consultants in our org have created for a customer. What immediately caught my eye was that a logarithmic scale is being used with a bar chart, leading to some really skewed data relationships. I felt like two things were wrong with it: 1)log scale with a bar chart, but 2) log scale using the raw numbers, not a log(10) of the raw numbers. I am fuzzy on my 2nd point, and would appreciate input on how to graph numbers to a log scale.
When I questioned this use, it was pointed out that frequently they have to display counts that range from 1 to 1 million and a log scale is the only way to do so.
Can I say anything to convince them besides, don't do it this way? Am I right to say that to even start using a log scale you must convert your data into log base 10 numbers?
I would like to point out the error while also helping them with an admittedly tough problem, mapping a scale of data that can go from 1 - 1 million.
Hopefully my chart example shows up here:

Forum: Examples of bad graphs and dashboards

Student Performance Dashboard using BIRT

Wed, 08 May 2013 19:50:25 GMT

Last year I was challenged by my manager to build the winning dashboard from the competition last year in BIRT (Disclaimer: I am an Actuate employee) as Stephen Few was presenting at our customer days. As one would expect I was not given much warning (asked Monday night, needed to deliver by Wednesday). As I always love a challenge of course I said no problem. Below is a link to a live report and an image of the result.

http://poc.actuate.com/few/winner

Now that Stephen has published his own solution I thought I would build that too (was a bit more challenging). Again below is a link to a live report and an image of the result.

http://poc.actuate.com/few/stephen

The dashboards are also interactive. One can sort the data:

Please let me know what you think and if you have any questions or comments.

Forum: Dashboard design

2nd edition of Information Dashboard Design

Thu, 02 May 2013 20:13:25 GMT

Do we know when the second edition of IDD will be released? I read somewhere (lost the reference) that it is due in the first half of 2013. I would like to know if we have a more precise date. Thanks.

Forum: The book "Information Dashboard Design"

Can eye-catching infographics and good data visualisation be combined?

Mon, 29 Apr 2013 07:45:18 GMT

Hi All,

I have collected a set of data which I think will be of interest to a number of people. I have started plotting the data using various techniques and I am happy that this reveals some interesting stories.

I am debating the overall style that I will use for this presentation. Currently infographics seem to be very popular - it would seem that tacking the term 'infographic' onto the end of a report will guarantee more attention. The problem is, 'infographic' seems to be synonymous with poor data visualisation.

I am wondering if the two can ever go hand in hand? It reminds me of this circle-lust post from a while back:

http://www.perceptualedge.com/blog/?p=799

The inforgraphic style imagery that was being discussed in this blog post was clearly terrible from a data visualisation perspective, but does look quite eye catching. The alternative, which Bryan suggested, is quite the opposite:

http://www.perceptualedge.com/blog/wp-content/uploads/2010/05/federal-spending-bar-graphs.jpg

Whilst good from a data visualisation perspective, it is very very dull!

The problem here is that infographics are designed to be eye catching, to grab the attention of the fickle reader, to draw them in and tell them a story.

Whereas data visualisation appears to be focused on end users that are already engaged and have an affinity with the data. Sales dashboards being a classic example.

So, my question is, can the two be combined? Has anyone seen an eye catching infographic that also follows the principles of good data visualisation?

Colin E.

Forum: Dashboard design

Bricks and trending

Thu, 25 Apr 2013 20:44:33 GMT

While reading about bricks, I was beginning to think if there is some way to include trending into bricks, while retaining the original quantitative value . For example by using gradient shading of the bricks (e.g. bottom row green, middle row yellow upwards into red) representing a downward trend, the inverse true for upward trends. Or perhaps using gradients in the bins which is programmatically more difficult. Obviously, by trying to implement trending into bricks, this would negate the ability to use diverging scales of positive and negative values. But if you were using all positive values – would recommend this approach of trying to implement trending values into bricks? Obviously your point #2 in What’s the Downside about border color would need to be taken into consideration. Wanted to know your thoughts on this.

Forum: Articles by Stephen Few

Building Insight with Bricks

Thu, 25 Apr 2013 17:07:38 GMT

In his April/May/June 2013 article, titled Building Insight with Bricks: A New, More Perceptible Method for Encoding Quantitative Values in Geospatial Displays, Stephen presents a new way of displaying quantitative data on maps, called "bricks," that he believes works more effectively than the usual methods.

What are your thoughts about bricks? We invite you to post your comments here.

-Bryan

Forum: Articles by Stephen Few

How can we make this better...

Fri, 19 Apr 2013 15:42:42 GMT

I mean other than the horrible colors. This is from http://www.fastexas.org/

The purpose of the site is to identify school districts that are spending their money very efficiently without sacrificing quality.

It's a plot showing school district spending vs "academic progress" (improvement, not performance)

Part of the problem is that High Spending is not seen as a good thing, so while you would expect to see "low" on the left and "high" on the right, that results in the best ratings showing up on the top left instead of the top right. We figured people were more likely to assume that top right was better than to really study the axes, so we swapped the bottom axis so it goes from high to low

The other thing is that we are not supposed to be assigning value judgements to which is better - high academic progress, or low spending. So in the second screen shot, we show 2 districts - one has excellent academics but spends a lot, the other is about average on both counts. Because the study is promoting efficiency, these districts both have 3-star ratings.

Since this is really confusing, after hours and hours of designing and redesigning and many many comps, tptb turned what was a nice simple scatter plot into this combination heat-map/scatter plot/infographic.

We are now redesigning the site and trying to figure out how to show this data in a more intuitive way, without assaulting the senses. We have some ideas that use kind of an off-the-wall approach, but I thought I'd post this here to see if we get better ideas from y'all.

Thanks,
RioBrewster

Forum: Examples of bad graphs and dashboards

Graphing with large ranges in data

Thu, 18 Apr 2013 09:20:39 GMT

I’ve got a bit of a quandary about how to display data that has large extremes. I’m frequently coming across data sets which includes sales and volume figures for a about a dozen different categories some ranging in the low thousands other in high tens of thousands. Trying to show these graphically clearly present problems – it’s nigh on impossible to differentiate and trends / differences in the smaller categories. Because the scales are skewed by the extremes.

The categories are set by the client (with no realistic opportunity to change until the project is renewed). The logical answer is to group them into more manageable chunks, but this will lead to considerable bloat in the qty of charts to present / deliver. Are there any other sneaky solutions that anyone has used before that will maintain integrity in the data (i.e. differences between categories, and showing trends within categories)? Just to note, the data typically includes 3 time periods for each category (i.e 12 categories, each with 3 time periods).

Regards

Whatx3
---

Forum: Graph design

Promoting bad graphs...

Fri, 05 Apr 2013 19:35:23 GMT

If there is an "anti-Christ" for bad graphics, perhaps the author of this article is it! :)

http://designfestival.com/5-ways-to-make-data-visually-interesting/

Forum: Examples of bad graphs and dashboards

Design Example 4

Tue, 12 Mar 2013 19:11:54 GMT

Hi Stephen,

Recently have found your website – really like it, thank you for very interesting and useful articles!

Going through samples of design ( http://www.perceptualedge.com/example4.php ):

Don’t you think that for Employment Candidate Review applying Color scales would be useful?

Especially if number of candidates is high:

Very easy to detect best performers in each category as well as overall.

Forum: Graph design

Finger Friendly software

Mon, 11 Mar 2013 15:01:53 GMT

Hi folks,

We're looking to create some moderately interactive reports and dashboards that need to be displayed and interacted with on a tablet (likely an Android tablet).

The thing is, I've not yet seen any packages that allow you to create truly finger friendly, drillable dashboards / reports very easily. Are there any good ones out there?

Thanks.

Forum: Visual data analysis software

Where to even start?

Sun, 10 Mar 2013 16:40:34 GMT

One of the clients of the company I work for is considering using a third party to produce their online dashboards. Looking at the example site (http://93.174.138.210/DemoApp.Web_deploy/demo.aspx#/Home), I can't even begin to explain why this looks like it'll be a very bad idea. Anyone want to help me compile a big list of reasons?!

Forum: Examples of bad graphs and dashboards

Feedback on iDashboards

Fri, 08 Mar 2013 16:21:45 GMT

I would appreciate any kind of practical experience with a tool called iDashboards. We have been offered this sw for evaluation and at a first superficial look it seems to give a good impression. The objective is to see whether or not to shortlist this application for a pilot implementation (Proof Of Concept).
Thanks very much.
Franz.

Forum: Visual data analysis software

Wealth Inequality - Wealth statistics in US presentation

Tue, 05 Mar 2013 08:09:17 GMT

Just saw this and thought it was some good visual narrative:

Forum: Visual narrative

sometimes high is good, sometimes not

Thu, 21 Feb 2013 18:13:50 GMT

I have a report, that will have multiple graphs on each page. In some cases a high number is good, in other cases a high number is bad. For example: Total cholesterol & LDL, a high number is bad, but HDL cholesterol is good to have more, what is an efficient way to show these kind of things within the same report and not confuse the user? Also, how would you incorporate measures over time into such displays?

The second issue I have is that I have a "score", which indicates the amount of risk a person has. The higher the score the lower the risk, everything I have played with tends to be confusing.

I have the books if you would want to direct me to something there.

Forum: Graph design