August 29, 2011

Overcoming resistance & building champions - Albert Birck, Maersk Line - as presented at ATUG

The August Atlanta Tableau User Group meet-up featured Albert Birck from Maersk Line as a special guest speaker, all the way from Europe. Albert gave a wonderful, humorous presentation that focus on:

Overcoming Resistance in BI Projects
  • Outgunned and outnumbered
  • Feeling like an outlaw
  • No funds, only problems to solve
  • Locked-in
And how 'guerilla-analytics' can help you deal with it

If you have a few minutes please participate in Albert's Performance Management survey done together with Henley Business School. The results will be available free of charge and he expects it to be very useful also for how you use business analytics to support performance management projects:

https://www.surveymonkey.com/s/perf_mgmt

Thank you, Albert, for sharing your insights and time with us.



August 26, 2011

Tableau Tip: 7 easy steps to create a combination chart with overlapping bars & a line

As a follow up to my previous post, which showed how a dual-axis chart looks in Tableau (compared to the Excel version I wrote about in the post previous to my previous post), I was asked by someone named Anonymous (I can never seem to identify him/her) to create similar instructions for building the dual-axis overlapping bar and line chart in Tableau.

Here are the steps I used to produce the chart (there are a billions ways to skin the cat, so take it for what it is):

Step 1: Place the Week dimension on the Columns shelf and the Measure Values measure on the Rows shelf

image

Step 2: Drag the Number of Records pill off of the Measure Values shelf to remove it

image

Step 3: Drag the Units pill from the Measure Values shelf to the right edge of the chart (i.e., the secondary axis).  You know you’re on the secondary axis when you see the dashed vertical line

image

You’re entire canvas should now look like this:

image

Step 4: Right-click on the Units pill, choose Mark Type => Line

image

You’re viz should now look like this:

image

Step 5: Right-click on the Measure Values pill, choose Mark Type => Bar

image

STOP!  Pop the top off that next bottle of Blue Moon.  Double-check your work.  You’re viz should now look like this.  If not, click the Undo button until you get back to a good spot.

image

Step 6: Drag the Measure Names dimension onto the Size shelf

image

Step 7: At this point, the bars are stacked.  So on the Analysis menu, choose Stacked Marks => Off

image

That’s it!  You’re done!  There’s some formatting you could do like adjusting the width of the bars, changing the number format on the axes, etc., but the essence of the chart is now complete.  Seven simple steps in 15 seconds (I wrote 90 seconds in my last post, but I timed myself this time).

image

Tableau: Dual axis chart with overlapping bars and a line

What took me a couple hours to figure out in Excel (see previous blog post) just took me about 90 seconds to create in Tableau (though I do have 4+ years experience with Tableau).
Notice that the bars are properly scaled now too.  The gray bar now longer is at $150 not $190 like it looked like in Excel.  Bottom line, use Tableau.

Excel: How to create a dual axis chart with overlapping bars and a line

I came across a challenge last week while working on a project.  I needed to create a dual-axis chart, with two bars on the primary axis and a line on the secondary axis.  Tableau makes this task incredibly easy, but I needed to do this in Excel.  Well, I really didn’t HAVE to create the chart in Excel, but others needed to be able to update the chart and they, gasp, don’t have a Tableau license.

It’s easy enough to create a dual-axis chart in Excel, if you want the bars side by side.  However, I needed overlapping bars.  There’s no standard chart design within Excel to accommodate this, which meant I had to come up with a workaround.  I perused the internet and didn’t find anyone else that had done this (I’m sure people have done it, but haven’t shared their work), so I wanted to share it with anyone that may need to do it in the future.

A sample Excel workbook can be found here (if your corporate network blocks Dropbox, send me an email and I’ll forward it to you).

The final product looks like this:

image

Steps to reproduce:

  1. Highlight your data, insert a 2-D clustered column chart

    image
  2. Change the bar and line colors if desired
  3. Right-click on one of the bars that you want on the secondary axis and choose Format Data Series
  4. Change the Plot Series On option to Secondary Axis
  5. With the bars on the secondary axis still highlighted, from the Chart Tools Design menu, change the Chart Type to a line

    image
  6. Right-click on one of the bars that are on the primary axis and choose Format Data Series
  7. Change the Series Overlap to 100%

    image
  8. The two bars on the primary axis now completely overlap each other.  You’re almost done!
  9. Remove the gridlines (important…I’ll explain in a bit) and add the axis labels (I hate it when people create dual-axis charts in Excel and don’t add the axis labels!)

    image
  10. Give the bars the “overlap” look (these are the most important steps to give the bars the proper) by right-clicking on the gray bar (i.e., the bar that’s in the first column of the table) and choosing Format Data Series
  11. Choose the Border Color option, select Solid Line and set the color to the same color as the bar (in my case, light gray)

    image
  12. Choose the Border Styles option, change the Width to 10pt (or whatever floats your boat) and change the Cap type to Flat and the Join type to Miter

    image
  13. That’s it!  You’ve create a dual-axis chart in Excel with overlapping bars on the primary axis and a line on the secondary axis.

    image

Now wait, I mentioned earlier that you should delete the gridlines.  This is critical because we’ve changed the height of the bar in the back by adding the border.  This leaves you with a couple of options:

  1. Make the border of the bar in the back really thin, but this makes it challenging to see the overlap
  2. Change the data to account for the thickness, but it’s kinda scary to me to alter the data because if you mouse over the point you see the original data
  3. Leave the bars thick so that you can clearly see the overlap

I would go with option 3 because the naked eye won’t be able to discern the difference to the bar height anyway.  The purpose of the chart, after all, is to give the overall trends and comparisons anyway, not focus on the original data. 

If that “change” in the bar size really bothers you, then simply add the data table under the chart:

image

Don’t we all love challenges!  I know I do.

TCC2011 – The Holy Grail of Strategic Decision Making (You may know the speaker)

At the Tuesday morning breakout sessions (11AM – 12PM) at this year’s Tableau Customer Conference, be sure to check out “The Holy Grail of Strategic Decision Making”.  For the second time in four conferences, I have been given the honor of being a customer speaker.  I’d love it if we packed the house; I think you’ll come away entertained and having learned something.

Summary:

Four-years of Tableau experience led me to a major turning point in my career, one that involved working with "buyers" for our products. These major customers expect insights and analysis from us, but they never experienced the "unbiased" stories that I've been able to tell with Tableau. 

This session will dive into new ways to work with customers and partners to identify opportunities that have been almost nearly impossible to detect without Tableau. The impact will lead you to think about the data you have within your own organization as well as third-party information to drive insight into your business that will change the way you think about making decisions.

I look forward to meeting many of you this year and I hope to see you at my breakout session.

August 23, 2011

Business Objects Evaluation circa 2008

Note to readers:
I've been asked by many people over the last couple of years to publish my evaluation of Business Objects back in 2008.  While SAP Business Objects has changed quite a bit since then and addressed some of the issues I raised, many of the critiques, evaluations and recommendations remain.

Enjoy!
Andy



August 19, 2011

Chartbeat – A cute but deficient web traffic dashboard

While reading an article on Sports Chart of the Day, a cute little gauge on the right sidebar titled “The Engage-O-Meter” caught my eye. Naturally I clicked on it (which must mean they did something correct with its placement).  The gauge directs you to a chartbeat.com dashboard that’s capturing real-time site traffic for businessinsider.com. 

The amount of information on the page is impressive, overwhelming and disorganized.  Two pieces of the dashboard confused my straight away (and they’re the only two sections of the dashboard you see when the screen first appears because there’s so much spaced used at the top for basically nothing). 

Let’s dissect these two charts/graphs.

What mistakes have been made?

  1. It takes up way too much space on the dashboard.  There’s a ton of white space in this one chart alone.
  2. The active visits number could be represented as just that, a number.  The gauge isn’t necessary and doesn’t add any value.
  3. The new & returning users are difficult to interpret.  Why do the numbers need to be so precise?  Would a relative measure, like % of total, be more appropriate?

I would present this information in a table. Ultimately we’re simply “looking up” the data, which lends well to a table.  We’re not attempting to determine any trends, which is what charts do well.

Active Visits

Now 8,342
30-day Min 3
30-day Max 17,971

 

Visitor Composition
New 20%
Returning 80%

Onto the second chart.

What could be improved?  What’s confusing?

  1. This “Top Pages” section, is merely a list of the top pages and their respective number of active visits.  Why not a simple list?
  2. Do the circles add any value?  To me they don’t.  Each circle represents 10 active users, but the layout of the circles doesn’t make sense.  If you count across, there are 19 columns.  Why not 20?  That’d at least make it easier to estimate the totals.
  3. If you multiply the number of circles by 10 in each section, you don’t get the same total as represented by the number itself.
  4. I believe, but can’t substantiate, that they’re making this similar to a Merimekko chart, which is a chart that encodes two variables: one for height and one for width.  I don’t see why this would be necessary at all.

Maybe representing all of the data as a bar chart would be better. 

 

In the end, with these two simple changes, I’ve been able to save a ton of space.  The rest of their dashboard could be reworked as well and maybe it’d all fit into one nice window.

August 18, 2011

ATUG Webinar Today 1pm ET - Albert Birck (Maersk Line): Overcoming Resistance in BI Projects

Join ATUG today at 1pm ET as special guest Albert Birck from Maersk Line joins us all the way from Europe to present his talk from the European Tableau Customer Conference:

Overcoming Resistance in BI Projects

  • Outgunned and outnumbered
  • Feeling like an outlaw
  • No funds, only problems to solve
  • Locked-in

And how 'guerilla-analytics' can help you deal with it

-------------------------------------------------------
To join this meeting (Now from mobile devices!)
-------------------------------------------------------

  1. Go to https://nsc.webex.com/nsc/j.php?J=715230391
  2. If requested, enter your name and email address
  1. If a password is required, enter the meeting password: stealth
  1. Click "Join"
  2. Follow the instructions that appear on your screen.

Meeting Number: 715 230 391
Meeting Password: stealth

-------------------------------------------------------
Audio conference information
-------------------------------------------------------

USA: 1-866-672-6771
International: +44 20 7198 1751
User code: 02 56 577#

August 12, 2011

Fertility Exposed – Tableau vs. Many Eyes

It’s been a while since I’ve checked out Many Eyes, but this one below has been on my list to review for a while now. Click around to get a flavor for their presentation. 

I’ve seen lots of vizzes done well on Many Eyes.  The viz above has many good qualities:

  1. The mouse-over feature is smooth and quick. I like the format of the text and how the state becomes outlined.
  2. You can easily switch between metrics with the drop-down box.
  3. You can show one, two or all maps at the same time to make comparisons more easily.
  4. Initially, when you show more than one map, the color scales are unique to each map, but there is a simple checkbox that aligns the scales.
  5. The brushing/highlighting feature works well when you show more than one map, which makes comparisons between individual states even easier to make.
  6. I like the choropleth maps, a feature we’re all still waiting for in Tableau.
  7. Switching from choropleth maps to States represented as bubbles can be accomplished in one click.  This would be another great feature to add to Tableau once/if they add choropleth maps.
  8. I particularly like how you can crop Alaska and Hawaii to make the overall map bigger without sacrificing those States.  It’s a bit annoying that you can’t do this directly in Tableau, however, you can work around it in a dashboard.

However, there are some improvements that I would make (and attempted to below):

  1. There’s not enough variance in the colors, making small variances between States harder to distinguish.
  2. While the choropleth maps are nice, they should be used with caution because the larger States can easily stand out simply due to their size (e.g., Texas)
  3. For some reason, most of the States are labeled, but not all of them.  Not sure why that is especially considering the unlabeled States in the Southeast.
  4. Many Eyes limits you to one view (i.e., a map or a line chart, but not both).  In other words, you can’t create a dashboard.
  5. If you view the map as bubbles, you can’t add another dimension or measure as a color.
  6. The default bubbles sizes make variances between States indistinguishable.  There’s a scale on the lower-left, but it doesn’t automatically adjust to the data in the view.  The only measure that is well represented by the size is the number of births.

Here’s my take, using Tableau:

With Tableau I’ve been able to make many of the improvements recommended as well as add vizzes that allow for quick insights:

  1. The map incorporates both color and size, which are set based on the measures picked on the right.
  2. Those same parameters update the other three charts.
  3. Brushing a State(s) in any chart highlights the State(s) on the other charts.
  4. Hovering over a bubble or bar reveals the details of the stats chosen.  The stats update based on the selections made.
  5. You can easily see that Utah has a small population, but the highest fertility rate.  Maybe it’s their religion? 
  6. I’m able to use the color scale consistently through all of the charts.

I made this viz with Tableau 6.1, which now allows hyperlinks on images.  This certainly simplifies directing people to your social networks and data sources.

How would you improve my viz?  Please share.

Join us at the August Atlanta Tableau User Group Meet-up!

Click here to register

Join us at the Atlanta User Group August 18th from 1-4pm @ Norfolk Southern.

Agenda:

  • Introductions & Networking
  • Overcoming Resistance in BI Projects - Albert Birck, Maersk Line
  • Break / Networking
  • Team Activity
  • September Meeting Discussion (any volunteers to host?)

This will be a hands-on meeting so please bring your laptop with Tableau.

Who should attend? Tableau User groups are open to any Tableau customers and enthusiasts.

August 2, 2011

Tableau Tip: Local vs. Global Filters on a Dashboard

A few weeks ago I chimed in on a discussion on Tableau’s forum regarding a concern/problem/feature request that a user wanted to see in Tableau.  Here’s a synopsis of the discussion:

Having to create duplicate data sources & bookmarks is burdensome, inefficient, and painful. 

Global Filters needs this functionality soon:

  1. Moveable Quick Filters – moving on any worksheets moves the global quick filter on all worksheets
  2. Ability to select or unselect multiple global quick filters at the same time
  3. Ability to turn On or Off across all worksheets with one button toggle
  4. Ability to make local to only specific worksheet(s), while leaving the filter for all the other worksheets
  5. Dashboard filters, filters that only effect sheets on the local/current dashboard

I’ve run into this myself several times and developed a bit of a workaround that I now use quite often.  In essence, instead of using a global filter (which impacts all sheets including and beyond the current dashboard), I create additional worksheets, but make them look like a quick filter.  Here’s an example:

Notice the three tables on the right.  These are actually worksheets.  I’ve created actions based on what you click in these three tables.  This method works for creating a filter that’s local to the dashboard.  It doesn’t completely address the issue raised, but it comes pretty close.