Time for a quick experiment on data visualization. Search Google for “data” and look under images. What you do not see is almost as instructive as what you do see. You won’t see tables and tables of numbers lacking context. Instead, you are treated to all sorts of graphical and pictorial interpretations of data, some almost reminiscent of an iconic screen shot view from The Matrix movie. We intuitively know, and the internet confirms, that reams of data, even if they are organized, are of very limited value without a humanistic visualization.
Computers are numerical, humans are visual. Thus, while computers are great for data collection and analysis, when it comes time to actually understand and act upon the results, it is time to bust out the charts. Data Visualization is the first step in Big Data success. After all, a picture is worth a thousand words!
To really understand the power of data visualization, its best to think about it from the opposite direction. Consider a piece of art. For the sake of universality say the Mona Lisa. We could break the painting down on its horizontal and vertical axis, assigning an alphanumeric designation to each coordinate point. Once we’ve done this we could describe, in near perfect detail, the RGB composition of each coordinate point. We would then have a large dataset describing the painting in granular detail.
While we would have a complete dataset that would give us a full understanding of the object it seeks to describe, anyone who doesn’t think something would be missing from this is not being honest. The visualization itself can be interpreted much more easily by the human than the gigabytes of RGB data points (although, we’ll find that these will have value as we advance through our analytics journey).
In the same way, many companies are still “flying blind”; having collected large datasets that may even be of acceptable veracity, but they just have not created an adequate visualization that provides the insights they need for interpretation and informed decision making. This would be akin to recognizing that you need some valuable information for your Mona Lisa, but not selecting paint and canvas to visualize it with.
The Sum is Greater than its Parts
Even if we forego the exponentially greater efficiency of gleaning insights off a chart rather than sorting through thousands of data points. Just like the painting is greater than the summation of its constituent parts, the profundity of seeing the insights visually with a chart is greater than discerning them from numbers.
In a way, communicating in this way with charts is an art.
This is the difficult work of data visualization – what is the best way to do it? Looking at a pile of data, the most effective visualization tool is not always obvious. Most everyone has been familiarized with the basics – bar graphs, pie charts, even histograms and Pareto Charts. However, choosing the best tool for the job requires a keen data knowledge. More often than not, the data would fit well enough into a few different types of charts. And occasionally the correct approach is to use more than one chart, allowing the different methods to supplement one another.
This ability to read the data and then know how best to present it is a learned skill. It requires an in depth understanding of the benefits and drawbacks of each data visualization method, the ability to “see” the data even when it is just a set of numbers, and an intuitive understanding of how humans learn.
While data visualization is a powerful way of not only gaining understanding from numbers, and also of sharing that understanding with others, it does come with one serious liability. For all the talk of numbers and data, visualization is a qualitative endeavor. Yes, the charts have numerical axes, but this does not come along with all the assurances of an exhaustive quantitative analysis.
Because of this, visualizations do not come with one of the most important pieces of an effective analysis – an estimate of the probability of error. Later when we discuss more advanced analytics including full statistical analyses, we’ll find that the chance of error may be determined (or even defined ahead of time!). If the organization makes this decision, we can predict that an outcome will fall within these parameters with a known level of confidence. After all, you will want to minimize the chance of making a wrong decision!
This is something a chart is not able to provide you. A chart simply shows a bell curve or a series of bars. While a very useful start, it is not all encompassing and warrants further analyses.
Not Just Graphs and Charts
Considering the art world one final time, it is an interesting sign of our abilities to gather data and visualize it in ever more interesting ways that artists have now begun to experiment with data. While the dataset from the Mona Lisa may not be much to look at, some artists find ways to make data beautiful.
For instance, taking self-tracking data from her personal FitBit, artist Laurie Frick created a two story tall aluminum panel display of her walking patterns. Others have created similarly unique visualizations out of obscure data sets.
There are no hard-set rules that says we must use bar charts and graphs to present our visualizations. Whatever best conveys the message the analyst wishes to put forth (while doing so in an honest way) is a good visualization.
Of course, visualization is only a small piece of the whole puzzle. Next week, we will be discussing analytical decision making. Visualizations inform this process and play a crucial role in arriving at a decision, especially when that decision is influenced or made by those without a deep background in data. Providing digestible data is crucial to decision making. Be sure to follow up next week to hear how BIG connects this with the next step of the process.