BeyeBLOGS | BeyeBLOGS Home | Get Your Own Blog

Main | December 2007 »

August 7, 2007

Visualization

Admittedly, I'm a picture-person. By that, I mean that I tend to think about things using spacial relationships; and I've been this way for a long time. In grade school, I can remember thinking, "I don't remember exactly what year Wyoming became a state, but I know that I read the answer about half way down a left-hand page somewhere in the middle of the chapter." I couldn't remember the detail, but I could recall the context in which that information can be found. Perhaps that just a sort of photographic memory... with astigmatism.

I know that other people think in different ways. My wife, for instance, reads a sentence (or more) at a time. It took me a while to understand that... "by a sentence, you mean one word at a time trailing into sentences?" "No, I mean I look at the page and see phrases and sentences, not single words." Wow. That's so different than the way I read. Maybe that says something about my intelligence, or maybe it just says something about how different people consume data in different ways.

I have a colleague that I've worked with for some years who likes reports with lots of numbers on them. Now, I'm a bit of a math geek, too -- I got a shirt from my daughter a couple years back with this on it:

01000100
01000001
01000100

Still, I think there's more value that can be derived from that sheet of numbers than can be delivered in the bare numbers themselves. Choosing the right pictorial representation can be very hard, but it can also be very powerful. Here are a couple of examples of relatively simple (and underused) visualization techniques that I think are very powerful.

Radar Charts:
If you plot two or more series on a single radar chart (or spider chart) you can deliver information about both which of the series "covers the most area" as well as in which dimension each series "out performs" each other series. So, you've got both point-by-point comparisons as well as an overall comparison available in a single view. The numeric equivalent might be detailed rankings and an overall average ranking.

Bubble Charts:
This kind of chart also allows you to deliver more than one kind of information in a single view. Bubble charts provide for both a horizontal and vertical plot as well as a third measure that is conveyed using the relative size of the marker on the chart. So, you could plot "number of infections" versus "number of flu shots" for a set of regional clinics and use the size of the marker on the chart to represent "number of nurses on duty" -- if you had a reason to think some pattern might come out of those three metrics.

Gantt Chart:
Don't think I'm just referring to project plans, dependencies, and resources here. Gantt charts can be used convey more information than that. A gantt chart, after all, isn't much more than an "activity" versus "time" chart anyway. Imagine a chart with patient volume on the Y-axis and time on the X-axis. Great. You know how patient volume changes throughout the day or over the year. Add a set of gantt charts to the bottom showing holidays, the school year, and other special events and your chart now allows you to correlate patient volume with various other pieces of information other than "time" generically.

You can imagine adding more dimensions to each of these kinds of charts using shape and color. You could add animation over time to show a trend. You could add sound as another mechanism for delivering information, reflect an upward or downward trend, for example. Three-dimensional charts are fine, but imagine a 7-dimensional chart that uses X, Y, Z, color, shape, size, and animation to convey so much more information. Clearly there's a practical limit to how much information someone can consume from a single view of some piece of information, though.

Still, reports with lots of number on them probably isn't enough to justify a multi-million dollar reporting tool any more.

Share: del.icio.us Digg Furl ma.gnolia Netscape Newsvine reddit StumbleUpon Yahoo MyWeb  

Posted by Paul Boal at 7:45 PM | Comments (0)

August 4, 2007

Pushmi-pullyu

Most of the data warehouse related projects I've been involved in that are related to a transactional system implementation, migration, or enhancement have followed the implementation of that OLTP system rather than lead or been concurrent with. I believe that's primarily because the data warehouse was also often the "reporting database" for those implementations. Part of the reason was to help justify the construction of the data warehouse - "We can off load reporting from the transactional system to the DW! The BI team can create those reports for people!" Very honorable, indeed, but those kinds of projects are often fraught with scope and requirements changes as the transactional system is implemented and people figure what it really means in the world of day to day work.

I'm in the middle of a 5 year initiative that is gutting and reimplementing the underpinnings of a all of the operational systems of a 20,000 employee organization right now. This is happening, though, because the powers-that-be in my organization decided it was part of their mission to become an "information enabled" organization. Besides implementing new and standardized applications across the company to ensure that we could collect lots and lots of information, the leadership also create a Performance Management department, which funds the Business Intelligence and Data Warehousing group. Insightful, I think.

Because Business Intelligence has such a prominent role in the organization, the implementation of reporting functionality (operational and some analytical) is happening alongside the implementation of the new ERP and billing systems. What's even more insightful, I believe, is that in some cases, there are Performance Management objectives influencing how the new systems are being rolled out - specifically in the are of master data management and conformed dimensions.

The theoretical concert between operational changes and business performance management makes a lot of sense - someone with great analytical modeling skills looks at data and determines that the business could perform better if some business process were altered. So, the business process is changed... sometimes by implementing a new application system that can help make people more productive... but that also means the business model and underlying data on which the original analysis was build changes... and the analytical model breaks. So, the data warehousing group waits for the application to go-live so they can rework their technical infrastructure to get the data and put it in some new set of data structures for the analysts to build new models on top of and go through the next iteration of process improvement. Those cycles take years to happen.

Wouldn't it make more sense if the operational applications that are built to improve business performance used a strategic analytical model as an input into what the system should look like and what kind of data it should capture and deliver back to folks doing performance management? Would it make any sense for the data warehouse to be built before the operational application, and be used as a validation that the new operational processes being implemented actually match what was designed to improve business performance? Once a business has gone through an initial cycle of process improvement, who should be pushing whom?

Share: del.icio.us Digg Furl ma.gnolia Netscape Newsvine reddit StumbleUpon Yahoo MyWeb  

Posted by Paul Boal at 11:15 PM | Comments (0)

August 3, 2007

Walking on Coals

With the second phrase from REM's Exhuming McCarthy, "walking on coals," try not to focus on the science of fire-walking, but rather the psychology and imagery - the idea of walking on coals as a test of mind over matter.

In the world of Business Intelligence and Data Warehousing, here we sit with gobs and gobs of data (rarely enough or exactly the right data, it seems, but gobs none the less). Here we sit with the world in front of us represented in tables and records, entities and attributes, facts and dimensions. The work it takes to derive value from that mound of information is very much a mind over matter challenge. Often, it looks simple on paper, but as you dive into the complexities of reality and start revealing "exceptions to rules" and "obscure business logic" and "poor quality data," it takes real cleverness and fortitude to turn that information into action.

Maybe this wasn't Michael Stipe's intent, but it works for me.

Share: del.icio.us Digg Furl ma.gnolia Netscape Newsvine reddit StumbleUpon Yahoo MyWeb  

Posted by Paul Boal at 7:45 PM | Comments (26)

Sharpening Stones

You'll notice the reference to REM's lyrics to Exhuming McCarthy in the name of this new BEYEBlog. I have to give full credit to my wife for choosing that inspiring idea as the title for my blog. I think it's incredibly fitting to describe how Business Intelligence and Data Warehousing fit into the world today.

Consider the first phrase "Sharpening Stones." The image I get in my head is of Neanderthals in the stone age chipping flat rocks into point spears that they use to hunt mammoths and saber-toothed tigers. It took tens of thousands of years to get from there to metal spear tips. Lucky for us, technology moves a lot faster. We've gone from the mere idea that a computing machine could be built in the 1800s to incredible small and powerful electronic devices in only 200 years. Still, I think that the information age we're in today is closer to the stone age than the bronze age. The computers that we build are giving us the ability to create and gather information in a way that we couldn't have ever imagined before.... but we're still barely figuring out what to do with all of that information that technology allows us to collect.

We're still just sharpening stones. We take a "collect everything that'll fit on the biggest hard drive approach" to information hoping we find something valuable. We don't have the precision and power of metalworks.

Share: del.icio.us Digg Furl ma.gnolia Netscape Newsvine reddit StumbleUpon Yahoo MyWeb  

Posted by Paul Boal at 7:30 PM | Comments (1)