What is User behavior prediction? Instead of Human behavior prediction, we won’t predict if the person wants to hug or kiss another person. Instead, we’re interested in their actions only in the scope of our application. Specifically, we want to predict why the user opens an app or specific screen, what he wants to see and suggest him to proceed directly to his desired outcome.
The most bright example that possibly every person with iPhone uses is the “App suggestion”
App suggestion by Siri in iPhones is really great thing. What actually bad is that that’s the only thing that predicts your next step on iPhone.
Literally no apps implement this ability. But they should. Take, for example, Evernote. Evernote has notebooks that act like categories, and then you have notes itself.
Let’s say you’re in the gym right now, you just want to open a notebook that has your workout. It requires 3 steps. When you open your app, you need to click on the navigation drawer and then find your notebook.
As you know, the main target of apps without the ads shouldn’t be user retention inside the application, but rather an ability of the app to solve the problem faster.
Firstly, we had very bad apps that were just able to handle the user’s problem. (Imagin Web 1.0/2.0). Now, we have apps that can handle your problem a lot better. We have computing power, we have great libraries and frameworks. Next big thing is to make apps working even better at what users want to solve, and essentially become the user’s secretary rather than a phone and calculator together.
That’s where our User Behavior Prediction comes in. It’s time for our apps to help us with what we want to achieve. And that’s where Machine learning comes in. Artificial intelligence showed itself well in different fields including character recognition, prediction of the load on servers and rebalancing them, AWS uses AI for phone verification(really nice thing). With powerful enough mobile phones we can even use machine learning on devices, that’s also called Edge AI.
Taking our previous example, if Evernote can integrate such technology, User prediction Behavior can show you a suggestion to open the required notebook or even note itself!
As a person who developed such a library for Android. I can easily say that it’s really easy to implement. The library makes it really flexible in usage, but if developers okay with writing it themselves it also won’t take too much time, and, frankly, easiest implementation in Java can fit under 10 lines of code, not saying about Python.
So why nobody uses it yet? As you know, there’s such a graph called “hype cycle”. User Behaviour Prediction tightly depends on Edge AI. Down below you can see that Edge AI just barely passed Innovation Trigger and starts its way in Inflated Expectations.
It makes User Behaviour Prediction even lower than Edge AI. So, I think we should expect to raise for such a technology in the upcoming years when Edge AI can prove itself.
It’s more a matter of stability, reliability, and accountability. Evernote has thousand of thousands of users daily. They can’t just integrate something that’s not so reliable. And that’s not the worse example, they have “Context” that shows you related notes. Big companies want to be sure about technology before they can use. The good part is that we’ve seen many times when technology comes really fast into the enterprise world.
To sum up, I would say that such technology should appear in the near future, but not in near months. And, hopefully, next time when you’ll be in the gym in headphones, Evernote will suggest the required notebook for you.