In a world where context is king for marketers and their personalization efforts, machine learning can make a big difference.
Abe Taylor, technical director at ICF Olson, drove that point home in a recent webinar, “Getting Started with Machine Learning.”
“It’s not so much the words that we're putting on our site but the context and what we're delivering to our customers, what they're seeing and the personalization that they're getting across their sites, whether it's product or language," Taylor said. "That is what is really driving them to continue to engage with us."
Beefing Up Personalization
The personalization pressure is on for both marketers and developers. The latter must build a solution that has the capability to take advantage of a preferred analytics platform. Marketers, meanwhile, must identify and attribute points to their different personas and accurately target their content to those personas, then validate their personalization efforts with A/B testing.
This is where machine learning and predictive modeling can help. It can analyze and process the user behaviors and analytics gathered by systems and help move beyond traditional personalization techniques.
ICF Olson is a Microsoft partner that helps companies deploy and use the machine learning and personalization capabilities through Microsoft Azure Machine Learning Studio. Here's a snapshot of some of the personalization capabilities that can be achieved through Microsoft Azure:
Smarter Personas with Training Models
Taylor cited a hypothetical beer company that has targeted "Andrew" as an ale aficionado. Traditional personalization techniques can help discover Andrew's love of ales and his history viewing content on ales, so naturally, it targets him as a potential "ales" customer.
Going further with machine learning, you could take the analytics data and train a model. Users could go into an analytics platform and pull all the profile of data around Andrew and find matching personas. The data would be fed into a Microsoft Azure Machine Learning Studio training experiment. If you find good correlations for accuracy and everything is lining up, then the trained model is validating the approach.
The even better news? It gives you a way to validate the content that you’re tagging for that individual persona.
But what if you have a weak correlation? It means your data is too random which can sometimes be an issue when working with a group instead of an individual, Taylor said. You may need to identify sub-segments, and you can use an algorithm itself to identify users who do match the sub-segments or run a different algorithm.
“It helps you have a better match there and reevaluate the personalization of content that you have on your website,” he said.
Being able to customize a user's experience based not on a generic pattern or a persona but individual variance once was a pipe dream for marketers. But, you no longer have to imagine, Taylor said.
Using Azure Machine Learning Studio’s REST services allows you to analyze an individual user’s behaviors and personalize the experience just for them. This is not a persona grouping thing.
“We can actually do that in real time,” Taylor said. “So the first thing that we would do is create some custom code or use the platform rules engine in coordination with an analytics platform's API to pull the analytics on the current user’s likes and dislikes. And then you can retrain the model for that individual user with a web service call with an Azure Machine Learning studio project."
Learn how you can join our contributor community.