Why So Blue?

I don't know if you've noticed, but I love blue.

In the olden days, before the internet and according to my son, when I lived in a cave and cooked over a pit with the newly discovered fire, we did all our charts in shades of grey.  Because only the execs had color printers.   As a result, people were so excited to create some colorful reports for the big giant heads that things got a little out of control. Some of the reports were so horrific in color, that I think they may have caused color blindness.

As a result, when I was asked to use color, I used blue.  Blue is a reasonable choice - almost everyone likes blue - both men and women, and it's associated with calm and clarity. It doesn't conflict with other colors or make a statement. Also, if someone chooses to print your blueful report on a black and white printer, you can pretty much trust the shades of grey that will come out.

Then there's color combining.  I'm not very good at that and find it especially difficult when I'm to use brand colors. Things can quickly get out of control when you are trying to use color to imply significance or range and then have to add 2 or 3 other colors that have are not responsible for anything other than making the brand recognizable. Somehow, you have to let the reader know that one color means good or bad, but the orange chart over here is orange because we like orange, even though both charts represent the measure sales. Same goes for the pink trend chart - we like pink too.

I'd suggest that if you do use a lot of colors - then use one dark color (blue or black) as your indicator/highlight/alert throughout the whole dashboard and stay away from diverging colors on the rest of the dash. That is, unless you are good at color and enjoy the challenge.

Click to read Smithsonian post (image on right is color blind simulated)
I'm fascinated by color blind people - I can't image the world that they see. I recently read this post regarding the possibility that Van Gogh was color blind. The image on the right is what they think he might have been seeing.

Isn't that amazing? I find the original deeper and richer and wondered if his other pieces would look so different.

Can you imagine this sunlit harvest without red? It twists my brain trying to figure out what someone would be seeing that would translate into this.

Someone needs to invent color blindness glasses.

You can load images into this color blindness checker by Kazunori Asada that was used to produce the image above. This is incredibly handy for dashboarding. You can take an image of your dashboard and test if it still makes logical sense to a color blind person.

I decided to look at some map options for diverging colors, to see how they might look to color blind people.  Here's some standard diverging pallets and what they would look like to a person with color blindness. Note: the most common types are protanopia and deuteropia (approx. 5-8% men).

For the most common types of color blindness, red and green result in golden greeny browns and yellows. For me, the most disappointing is the temperature (blue/green/yellow/red) color palette. It really loses it's effectiveness under all lenses. I like what happens under the tritanopia lens, but the red is a bit startling. I don't dislike any of thse color combo's but they aren't what I would have expected if I was using the these combos.

Here's what happens with blues (sequential and diverging):

Notice I left out the red? It seems that there's not much point if it just turns into the same color as green and red has such a strong association with 'bad' or 'look out' that I rarely use it. And I'll probably never use it to make something stand out again after seeing this. I do like what happens with the gold/blue and green/blue diverging and will consider using those in the future.

What makes me blue?  When my work gets dismissed as 'pretty'.

I once had a badboss who regularly commented in meetings that once the dashboard was finished (in 6-8 months because this was a stack BI department), "...we'll give it to Kelly to make it pretty." My passive-aggressive response?  "Just make it blue."


5 Tips to Good Vizzin'

Before Tableau, I made a lot of dashboards in Excel. It has been quite difficult to free myself from the constraints that Excel (or other standard tools) forced upon me when it came to storytelling with data. Let me explain.

In the past, with other tools, I would envision the outcome first and work backwards.  Often I would be specifically asked for a certain type of chart showing certain information; sometimes I would be asked to investigate a problem, but I would still jump to the type of chart(s) I needed first and then go about getting my data into the state needed to put it in those charts.

If I was part of a committee or project it could be even worse - a dozen people drawing a dashboard on a whiteboard before anyone has even collected the data or had any understanding of data, analysis, or dashboard development. Disastrous. This video nails it - how many times have you felt like Anderson? That you're trying to explain that blue is not red.

Today, with Tableau, I go about it differently. I don't start with the assumption that I need a particular chart and I don't assume that I know exactly which variables to consider, but I have some hunches to start with.

It does get more complicated when you're building a dashboard that tells a story and isn't just a bunch of charts crammed together. Here's some basic, simple tips that hopefully will help.

1.  Know Thy Audience

Who needs to know the answer to the problem?  Just the VP?  Or will others need to use what you've built to make or track changes? Don't try to just impress the VP and forget that he/she will be sending this to others if she/he finds it useful. Will it help them?

How will they use what you build?  Laptop, mobile, live presentation?  How are they used to getting information?  If it's been static, printed reports, you will need to start slow and do a lot of handholding.

Keep it simple.  Less is more.  Is your audience used to dashboards or just seeing one chart?

Are they distrustful of anything other than tables of numbers?  Will they want to see all the data?  You could set up a dashboard that explains how to extract the data, providing them with basically a pivot table and multitudes of filters.  Also, remember to include an "About the Data" dashboard where you provide all the details about the source, extraction, inclusions/exclusions, calculations, and definitions.

2.  Know Thy Data

Get in, get dirty with your data.  Analyze the hell out of it. Understand the relationships between all your fields.  Deal with your nulls - know where they are.  Make a frequency table of all your numeric fields.  Know what all the categories are in your dimensions.  Replicate existing tables to test your results. Test your new calculations - make sure they behave properly when filtered or aggregated.

You need to know and understand your data's history.  Where did it come from, what's the work process behind it, who and when is it entered, what does it include, what does it exclude?  If you wrote the query that's great (don't forget to document), but if you didn't, you need to know all the details about it.

3.  Preguntas

Start with What and keep asking Why.  Exhaust your curiosity.  You may not use all of the charts that you create in this exercise, but you will have all the ammunition that you need to put the story together.

Put your questions in the Title of your views (charts and tables), this will help arrange the story when you put the charts on your dashboard. You can clean them up when you are finalizing the dashboard.

Remember your audience - what would be important to them? What questions will each chart illicit from them?

4.  Talk, Talk, Talk

That's 3 talks.

When building your dashboard, think in 3s.  Three charts, 3 filters, 3 sections, 3 colors, 3 fonts, etc. Less is great, but try not to go over 3 unless there is good reason to do so.  This is where you begin; until your audience gets accustomed to the style and interactivity, this is the simplest approach. It's a suggestion - not a rule.

Don't try to WOW them with everything, you will only overwhelm them.  Initially, you only have a few seconds to draw them in.  If people are overwhelmed or confused they are very likely to close your report in a few seconds and never go back to it.

Once your audience gets more savvy with this reporting style, you can build more complex dashboards.

Imagine your dashboard is in an elevator with the CEO and it has three sentences to back up it's story.  So make sure your charts are related - start high level, then go to detail - or show 3 key measures that impact something.

Simple is hard. Especially when you're in the thick of it. It's much easier to write a run-on sentence than it is to use proper grammar. It's easier to make 75 charts and stick them in a PowerPoint deck than it is to build a smart, clean, interactive dashboard. Remember this: Complexity is often used as an excuse for Obfuscation. Someone else may think that you couldn't get to simplicity and not trust in your results.

5.  Shuddup Already

Take some time away from your dashboard. At least overnight. Then when you open it up the next day, try to look at it from a bit of distance.  

Look at the chart that grabs your attention first. Is this the chart you want the user to focus on first? Does the next chart relate logically?

Is it screaming at you? Start to take away stuff. Unbold, make white space, calm the whole thing down.

Did it seem like I was screaming at you about simplicity? Sorry about that. I'll shuddup now. Happy vizzing!


Bubble, Bubble, Oil and Trouble

There's a lot of news about the Keystone Pipeline lately. Will Obama give the go ahead or not? Man, I wouldn't want to be in his shoes. This is not a simple black and white issue. Disclaimer: I'm a Canadian born in Northern Alberta and I have a lot of family members who work in the oil and gas industry or who benefit from it. I'm not for or against the Keystone Pipeline, I'm seriously conflicted.

If you want to learn more about this issue, check out Leslie Young's series on Global News. She's done a heck of a job getting this data and the series isn't biased. Another good source is the Guardian (always best to go to Britain for unbiased North American news) or our beloved David Suzuki (who is one of the most credible environmental activists on the planet).


New Year Revolutions

Good times and bum times, I’ve seen ‘em all
And, my dear, I’m still here
Flush velvet sometimes
Sometimes just pretzels and beer, but I’m here
I’ve run the gamut, A to Z
Three cheers and dammit, C’est la vie
I got through all of last year, and I’m here
Lord knows, at least I was there, and I’m here
Look who’s here, I’m still here
  - Stephen Sondheim
It has been a wonderful year. Nothing got broke that couldn't be fixed,
no one went to jail and no one died. The world kept turning and my family was funnier than ever.
This has been my standard for success since raising a teenager.


Finishing Touches: Fonts

If you are like me, fonts are the last thing you think of when building a viz or dashboard.  To make my life simpler, I decided to put together a few font combinations for quick reference.

I don't know much about font combining  (other than what I've read here), so I try to keep the number to a minimum and try not to have them competing for dominance.

This is pretty much the list of fonts I use in Tableau. I know they aren't exciting, but I can count on how they will behave on different Windows or Mac versions and different browsers. Besides, when it comes to dashboards, I don't want the font to be the star anymore than the colors. The font is there like any other element; to help convey the message that the data contains.

Please note that I'm talking about work, work here. When it comes to play visualizations, all bets are off and all fonts are fair game, even ComicSans.


Canada's Electoral Voting Districts

Canada's next federal election isn't until October 2015, so this viz is almost 2 years too early.  But I was so excited to receive the shape files for the electoral districts yesterday that I had to play.

Big, HUGS to Zen men Allan Walker, and Craig Bloodsworth at The Information Lab and Adam Riley at Altyrx and  who created Tableau Mapping Monday and the central Tableaumappingbi repository for the shape files they've created.  There's also a LinkedIn mapping group for discussions.


UBC Magicians

Click to visit site
Today I was invited to talk with grad students in a course on Information Visualization and Visual Analytics  as part of  the Magic program at the University of British Columbia (UBC).  I had met the Director of the program (and CiFER Researrch), Dr. Vicki Lemieux, at the Tableau 8 Roadshow here in Vancouver in the spring and she had asked if I would come speak to her class about working in this field.  Vicki is one of those truly precious professors who are not only experts in their field and love it, but also love teaching and are brilliant at it.  She explained how the class runs; she usually presents some new information and then the students work in groups to further explore and share their new found knowledge.

Magic stands for Media and Graphics Interdisciplinary Centre.  This is a great program which also offers students the opportunity to work with local business and government organizations:


Excel and Tableau: When a Table is just a Table (and not data)

There's all kinds of 'analysts', but perhaps the most prolific is the Excel Analyst.  That pretty much encomapses everyone who works with computers.  Excel has provided everyone with the ability to conduct a wide range of analysis and reporting for years.  Anyone can make a chart or table in Excel - you don't need an understanding or training in data analysis.  However, things can get confusing when we move to an analysis tool like Tableau.
Part of the problem is that most of us use the terms Table and Data interchangeably (because we can).  Here's the rub:
Tables, charts and maps are presentations of Data.
In relational databases, Data is stored in Tables.
So a Table can present Data or contain it.
The problem occurs when we try to take a presentation Table and use it as Data for analysis to make a presentation Table or Chart.  Confused?  Yeah, don't worry, lot's of people get confused by this.  It's why most Open Data is available as a presentation Table and not Data.