Yes, folks, I am back. Wait, didn’t I write that before.
Well, after having my 1st child, I spent many months (just shy of 10, to be exact), noodling business intelligence, and the concepts I had previously discussed on my blog. For the last 5 years, I have been touting the need for better search integration, offering up the BI mashup concept before people really understood what a plain vanilla dashboard was, and was met by glazed stares and confusion. Now that folks are catching on to the iGoogle experience, and the ability to “mashup” or integrated several points of interest or relevance into a dashboard, I want to discuss this topic again. But, this time, I want to apply the concept of the Decision Engine instead of just the Search Engine when it comes to ways to make BI content more meaningful, more relevant and more useful to end users.
Side note: “mashup” is still not a recognized word in the spell-check driven dictionary lists for the greater population of enterprise applications.
Coupled with my mashup passion was my belief in eye-tracking studies. Eye-tracking measures the human behavior of looking at something and measuring the concentration of the eyes on a particular area of a particular object of interest, say a website for example. In the case of business intelligence, I applied eye-tracking studies to the efficacy of dashboard design in order to better understand the areas where the human brain focused concentration vs. those ignored (despite what the person might say was of interest to them).
Advertisers have known about eye-tracking studies for years, and have applied the results to their business. For example, the eyes will focus on the top left corner first. Whether a TV screen, a book, a piece of paper or a dashboard. It is the area of the greatest concentration. Therefore, special importance has been paid to the piece of advertising real estate. And since the popularity rise of folks like Stephan Few of recent or Edward Tufte, whose design principles for effective dashboard design have driven many a BI practitioner to rethink the look and feel of what they are designing, this concept of top left is more important has become commonplace.
And, the handful of other book grade principles have risen to the surface too: less is more when it comes to color, overuse of lines in graphs is distracting, save pie for desert (pie charts, that is), etc. But tying it all together is another story all together. Understanding how human perception, visual perception and dashboard design meet is a whole other can of worms, and usually requires a specialized skill set to fully “grok” (sorry, but I love Heinlein’s work). 🙂
Excuse my digression…
Take a look at this image which shows eyetracking results from the three most popular search engines in 2006:
Notice the dispersion of color measured in the Yahoo and MSN examples vs. Google. This is correlated to the relevancy of the results and content presented on the page. And 4 years ago, Google’s search engine was a popular go-to tool for many when it came to finding related websites to help answer questions. Fast forward 4 years, and MSN is now Bing, and what was the search engine is now the dubbed “decision engine.”
The advent of the decision-engine in my eyes is because of the dilution of search engine effectiveness based on the flood of results presented to end users. In fact, I am sure the results of an eye tracking study from 2010 would be vastly different as a result of the exponential growth of web-based content available for crawling.
The same has occurred within enterprise business-intelligence platforms. What was introduced as powerful has really become inundated with content, in the form of reports, objects, dimensions, attributes, attribute elements, actual metrics, derived metrics and the list goes on and on.
Superficially, search was introduced as an add-on to the enterprise BI platforms. An add-on; really, an afterthought.
To the credit of the solutions on the market (grouped into a collective unit), people didn’t realize what they didn’t or better put, needed to know when building the technology behind their solution offerings. And they needed to start somewhere. It was only after BI became more mass-adopted in corporate America, and the need grew pervasive into even the smallest Mom and Pop shop for some level of reporting, that people began to realize that getting the visualizing the data was one thing; finding the results of those visualizations or data extractions was an entirely different can of worms.
At the same time as this was happening, the search giants started innovating and introducing the world to the concepts of real-time search and the “decision engine” named Bing. Understanding the statistical models behind how search algorithms work, even simplistically, understanding enough to be dangerous, is a key that any reader of this blog and any BI practitioner would be smart to invest their own time into doing.
In a nutshell, my belief? Applying those principles and eons of dollars thrown at optimizing said models (by the search giants) is an effective way for BI solutions at any level to leverage the work done to advance search research and technology, instead of just patching BI platforms with ineffective search add-ons. Just look back at the Golden Triangle study graphic above, and remember that long before BI design experts like Tufte and Few said it, advertising gurus knew that the Top Left real estate of any space is the most important space to reach end users. So, instead of thinking of search as a nice add-on for your BI platforms, why not see it as a necessity. if a report is loaded into a repository and no one knows about it, was it ever really done? Let alone meaningful or valuable enough to be adopted by your end users? Think about it…
Musings on ML, Deep Learning & AI 