I have always been interested in the information science field, specifically, discovering and mitigating extraneous web activity. Web analytics is a general term for the data needed to optimize how users interact with an online system. The data can come from tracking all kinds of actions, including web visits, clicks, time spent on a page, etc.
Before I started working with machine learning systems at StormGeo, I was working within education. Not as a teacher, but within learning analytics, which truthfully, is not too far off from what I now do.
Learning analytics, in basic terms, analyzes how students and teachers use any web-based tool. This data can help define which parts of the curriculum are wasted, how individual students learn best, or whether a student needs targeted intervention to succeed (before failing the test). This process of analyzing users’ behavior and interactions with a tool can be applied to any visual interface, including StormGeo products.
Some things we at StormGeo look into are: How often do our clients access our tools? How long do they spend during each session? How do they navigate within the system? By keeping track of how our clients use our tools, we’re able to see where there are potential issues and how long it takes to do the things they do most often. This insight not only helps us with product development (fixing known bugs and continuously improving our products) but also with creating new products and services – i.e. new tools that would make our clients’ experiences even better.
One example of this could be creating something called a collaborative filter within StormGeo systems. Netflix is a great example of a system that recommends things based on this filter, in that it creates personas based on the information being fed in from its users, then dynamically personalizes recommendations for each persona. So if you and I have watched the same ten TV shows, and you just recently finished watching Stranger Things, it’s likely that that show will pop up in my ‘Recommended for Sigve’ section.
Another somewhat newer example of this filter is smartphones, which have logged users’ behavior for quite a while, but haven’t started suggesting actions until the past couple years. For example, asking users if it should set an alarm for the time they normally wake up. The possibilities for relying on these systems to self-personalize is endless.
Recommendation filters can have a huge impact on any industry that tracks user preferences or how their systems are used. Within our industry of weather forecasting, there is great potential. For example, when StormGeo delivers forecasts to customers, we are able to gradually customize them to complement the usage pattern of each customer. On top of that, customers within the same industry can mutually benefit from the same pool of recommendations. Not only does this system help the customer find their preferred setup, it can also provide convenient improvements to their experience with our services. Ultimately, it’s about your ease of use.
This technology opens up new ways of providing decision support to exceed our clients’ expectations of what kind of personalization we can provide to them. It’s something we’re heavily looking into, because at the rate of technological development, what starts as innovation will soon become an industry norm for all.