There is an often-quoted figure that “90% of the world’s data has been created in just the last two years.” That sounds very impressive, but I will admit wondering, “so what?”. With so many ways for people to access and create data, the figure is hardly surprising. But that’s our new baseline- we have a lot of data. And the reason it matters is because now we can ask- and answer- questions in ways we could not before. That puts a lot of pressure on the one asking the question though…
Which is why I was reflecting on just how important it is to ask the right question! With all the data out there, and tools like Veera making it so easy to blend and analyze that data, you can assemble information on almost any area of interest. But that data only makes a difference if you’re being thoughtful about what you’re investigating. I want to lead off with an example.
I get to work with lots of folks from diverse industries- and from my work with those from Higher Education, I know they care a lot about student success. But what’s success? One team started out tracking whether students returned the following year. After they analyzed their data, built a predictive model for their initial y-variable, and checked their results (a process that takes less time than you might think), they reevaluated their approach. The models were helpful, but they needed to catch the warning signs sooner to truly help their students. They realized academic probation was the common thread for students who were dropping out and so they adjusted their y-variable. This gave them earlier notice, and more time to make a real impact.
This iterative process was enabled by fast and easy to use technology and allowed them to go deeper. They didn’t look at just student attrition, but what was causing students to end up on academic probation to begin with. Now they’re learning how to prevent students from even getting that far off track- and their improving their retention rates right along with their students’ success. This isn’t even the only option they had- another possibility I’ve seen is predicting first semester GPA, and even comparing that GPA against expectations. Now we’re talking about pinpointing at-risk “underachievers”, but also finding out what helps students “overachieve”. That’s what happens when you get creative with your y-variable. And it certainly helps to have tools that make the process faster and easier.
That’s just one example. Regardless of your usage, there is always value to thinking carefully about what measurable event fully reflects what you are interested in. Sometimes, you have to stick with an industry standard, like premature readmissions in a hospital setting. Other times, you have more leeway, and can explore new options. More examples? Here are two that really break the mold:
- Analysts at Target realized they could optimize mailers if they could predict which patrons were pregnant
- Impacting morbidity rates and saving money by reducing ambulance response time
With these as proof, always keep in mind that you can influence your overall goal through several different measurable y-variables. Target chose to predict the characteristics of mail recipients, rather than just predict response based on those characteristics. They got ahead of the curve. Jack Stout looked at ambulance availability through a different lens, and started a trend that has become a standard. In addition to the personal recognition, the nation has surely benefitted.
Part of creatively exploring the ways you can influence your overall goals is trying out ideas. Being able to iterate new projects is necessary, and usually prohibitive. But having ways to re-structure your dataset, incorporate newly related fields and save time on the statistical tests makes iteration possible. Our customers accomplish all of that with our tools. How have you tried to measure your outcomes? Have you wanted to try something new, but want to chat it out? I love to explore this idea of iterative analysis, so leave a comment or email me!