xAPI is not magic: V - xAPI correlates learning and performance

By jPablo Caballero | January 14, 2017

To correlate data from two different sources, you must first gather data from both sources.

Another topic that is sometimes mentiond in the context of xAPI, and that needs to be examined with critical eye, is the correlation of learning with performance.

This is in a way the holy grail of training. We want to demonstrate that training has a positive impact on real world performance. Some are quick to say that xAPI almost “solves” this problem. Again, xAPI can help -a lot-, but it doesn't per se solve the problem.

Most of the conversation around xAPI revolves around learning activities. After all, this initiative was born in the L&D sector. But the very flexible nature of the Subject + Verb + Object can be used for any type of activity, not only learning activities.

That's good, because to correlate learning data and performance data, we obviously would need to gather both types of data. The fact that we can gather and express both in the same ‘language’ will make it easier to analyze side by side and try to draw conclussions from the analysis. So, xAPI is a step in the right direction.

So, xAPI data tracking needs to get outside of L&D and be implemented in other apps, or at least, allow other apps to express their tracking data as xAPI statements through connectors.

xAPI has to extend well beyond the L&D space to many other areas and systems. For example, think of a sales person: we want to see if there's correlation between her behavior during online tranining, during classroom training, from the questions she asks in forums, the videos she watches in YouTube, etc. and how she handles clients, how many deals she closes and how much money those deals represent (real-world performance).

To do this, the CRM system, the “financial” systems that the company uses to track all this should be using xAPI as well to report all these “activities” (contacting a customer, sending and receiving email, closing a deal, etc.). Only then we will be able to run an analysis to find correlations.

Convincing other corporate departments to modify their systems to use xAPI (or to develop systems that brigde the original data from other systems to make it available as xAPI statements) is no easy task.

So, again, the fact that xAPI has the potential to pave the way so these analyses are feasible, doesn't mean that it's easy or that it's going to happen without some serious effort.

And of course it is necessary to always keep in mind the basic tenet of statistics and data analysis that “correlation is not causation”.