Author: Arijit Sengupta, CEO and Founder of BeyondCore
Many people consider visualization products like Tableau or data discovery products like QlikView to also be analysis tools. This is not true. While these products show the graphs you specifically ask to see, analysis tools figure out the right graphs to look at.
Analysis involves three main components:
- Asking the right questions (figuring out the right graphs to look at)
- Correctly interpreting the graphs (evaluating statistical soundness)
- Understanding the underlying causes of the observed patterns
In effect, analysis is about figuring out what is important and why it is important. Visualization, on the other hand, is about presenting what is important in the most visually attractive way possible.
The risk we run when we confuse analysis and visualization is presenting the wrong graph beautifully, highlighting apparent meaning where there is none.
People say lies, damn lies and statistics. What they should really say is lies, damn lies and graphs. Graphs lie, not statistics.
Let’s say you’re a hospital administrator looking for ways to lower costs while maintaining or even improving quality of care. You start to explore costly situations where the length of time in the hospital is unusually long. Much to your surprise, you find a huge variability between men and women in three categories.
Very interesting. So you dig deeper over the next couple days or weeks, and explore the drivers behind this. Maybe you even decide to act on your findings.
Stop here. From a seemingly interesting visual, you just unknowingly strayed from your cost-cutting mission. Take a closer look at these two graphs.
What the graph on the left doesn’t tell you is that most of this information is completely irrelevant. It’s just noise. When presented without a statistical context, a lot of time and effort may be wasted on this unimportant pattern.
On the right, BeyondCore highlights statistically significant insights, and greys out entire sets of bar graphs to signify their insignificance. Behind the scenes, sophisticated analytics take place, so not only do you have a great visualization with deep insight, you can immediately explore more details behind the dynamics that matter most.
||Goes beyond the visuals
||Not relevant: category 1 and 3
|Explore all three further
Only investigate category 2
Here is a further example of a visually stimulating graph with no statistical importance:
Had someone just seen this first graph, he/she may have spent unnecessary time studying a visually stimulating graph. BeyondCore automatically knows the information is trivial, and instead presents the information as such:
This graph has been automatically greyed out, and the y-axis has been automatically set to 0 so that the statistical irrelevance is visually obvious.
As both examples demonstrate, BeyondCore provides statistically relevant analysis, while ensuring users are not led astray by visually stimulating data sets.
Self-service visualization has become extremely popular. However, as demonstrated in the examples above, having business users without proper statistical training try to visualize data themselves is extremely risky. This is why the BeyondCore mission is to empower every business user with self-service analytics.
BeyondCore and SAP HANA
As part of SAP’s Startup Focus program, we are bringing the power of self-service analysis for business users to any data in SAP HANA. Any user can now analyze such data with a single click. This is the easiest way for a business user to analyze as opposed to just visualize their data in HANA today.