Back in the early 1990's, I created one of the first metric dashboards for the telecommunications company I had just joined to help improve business processes.  My goal was simple:  build a web-based dashboard that let management view charts showing the performance of key processes.  Sounds pretty simple, right?  Today it would be, but 20 years ago creating "business intelligence dashboards" was a convoluted process at best.  

Here's what the workflow looked like back then.  First, I created the intranet site that would serve as the holder for the dashboard charts.  It was pure Microsoft FrontPage with a black background, animated GIFs, and crazy colors — all the classic 1990's stuff.  I created these static dashboard pages for each of the key areas that we wanted to manage.  Want to see the key performance indicators for the installation process?  Boom — click the "Installation Process Metrics" icon and you're there.

The problem was getting the charts created.  The process wasn't exactly simple:  wait 30 days for the data to come in.  Use statistical software to analyze the data, highlight significant changes or outliers, and create the chart.  Add captions to the charts so that management team understood where problems were occurring.  Take the chart and text, embed them into the static web page and publish.  The charts had no interaction and any "drill down" required me to manually create a new chart for each possible drill path, but it worked.  Our team could see exactly where to focus their time and efforts.  We were able to avoid chasing down extraneous data points that looked scary, but were just random noise.  The problem was that it was a nightmare to maintain when you had to do this for each of the 30+ metrics used to manage the business. Oh, and the data was up to 30 days old — "real-time" analytics were just a dream back then.

Today creating analytics dashboards for performance tracking is simple.  Using one of the many tools available on the market these days makes creating beautiful, maintainable, and flexible dashboards simple.  The power offered by these platforms is beyond what I ever could have imagined back in those static web page days.  The dream of providing real-time insight into key business processes to executives has finally come true.  It's an amazing leap forward in use of metrics to manage a business.

This chart looks nice, but lacks the elements to tell if a problem is occurring or where to focus your attention

This chart looks nice, but lacks the elements to tell if a problem is occurring or where to focus your attention

Or is it?

With the emphasis that has been (rightfully so) placed on clean, concise, and beautiful analytics on dashboards, we've taken a step backwards.  Today's charts often look like blown-up sparklines, devoid of any information except for the data line itself.  And that's not a good thing.  With the amount of raw data available in easy to deploy dashboards, we risk being overwhelmed by pretty but useless charts — unable to separate out signal from noise.  We risk bombarding users with gorgeous analytics that leave them guessing about where they need to be focusing their attention.  Does it have to be this way?  Is it possible to provide a great user experience to dashboard users and still deliver critical insight into exactly where problems are occurring?  It's not only possible, it's already been done. 

Walter Shewhart (image courtesy of Wikipedia)

Walter Shewhart
(image courtesy of Wikipedia)

More than 90 years ago, a statistician named Walter Shewhart developed a technique for helping to separate signal from noise.  Called "control charts", these line charts had one primary purpose:  help users identify when a process or pattern has changed.  That is, they let you know when that little outlier of data is random noise or an actual signal that something has shifted.  In the years since their creation, these charts have been used for everything from identifying assembly lines gone awry to improving call waiting times at service centers.    These control charts were the analytics that we constructed by hand and used on our metrics dashboards back in those early days.  They saved us tremendous amounts of time by helping us understand when to act on a spike in a chart and when to ignore an outlying data point.  And yet, you rarely see these charts in use today.

A control chart circa 1996.  Ugly, but there's a lot of information packed in there...

A control chart circa 1996.  Ugly, but there's a lot of information packed in there...

Maybe it's because control charts, as they've been traditionally displayed, aren't the easiest to understand.  They are usually shown as two stacked charts and require familiarity with basic statistics and rules for interpretation.  Assembly line workers had to needed training in basic statics and in the interpretation of the charts they needed for their jobs.  But this doesn't need to be the case today.  Modern systems can alert a user when one of the statistical rules is violated, calling attention to the exact time when the issue occurred.  They can help focus the user's attention on just those items needing action separating out the signal from all that noise.  As even more data becomes available to management, marketers, and sales teams, the ability to determine when a real problem occurred and when fluctuations in that beautifully-rendered chart are just random noise will become critical.  

With all the buzz around predictive analytics that project what might occur in the future, we've put on blinders to patterns and shifts that could give us even more insight into our performance.  We've turned our attention to predicting the future and lost sight of all the rich information we can learn from the past.  The next generation of dashboards needs to take the concept of finding critical patterns and make it much simpler for the average user.  I hope to see this new breed of analytics perform the following:

  1. Build control chart rules into analytics platforms, but hide the mechanics from users
  2. Use the rules to mark patterns in the data showing where trends and shifts occur
  3. Allow users to annotate root causes
  4. Report on historical patterns so users can see the frequency with which underlying root causes occur

It's time to turn our attention back to the tools and techniques developed a century ago and bring them into the modern age for our dashboards.  We've solved the problem of getting data to users quickly, cleanly, and efficiently, now let's help them figure out where they should be taking action.   Applying the tools of the past in the dashboards of today is the answer.