Traditionally, machine-learning personalization has been used by e-commerce sites. As a refresher, machine-learning personalization utilizes algorithms and predictive analytics to dynamically present the most relevant content or experience for each and every visitor. As a consumer, you have probably experienced this type of personalization across the internet in the form of product recommendations. But most B2B sites are newer to machine learning, particularly for enhancing their demand generation and content marketing programs.
At Evergage, we’ve been adding machine-learning personalization campaigns to improve demand generation on our own site. In this post, I’ll describe a few of these campaigns to give you ideas of how you could use machine learning on your own site to increase engagement and drive more conversions.
1:1 Content Recommendations
Why did we do this? When someone lands on our site for the first time, we want to provide them with the most relevant experience based on all the information we can learn about them. This often starts with the information we can detect about first-time, anonymous visitors. We typically identify their company name and industry and tailor their experience based on this. Then we evolve the experience as each visitor engages with the site in the current and future sessions. The more information we gather about a visitor based on his behavior, the clearer the picture gets of the person’s true intent and interests, which is key to the accuracy of the machine-learning content recommendations.
We have an area on the homepage that uses machine learning to recommend a relevant eBook to each visitor:
This 1:1 recommendation is driven by a collaborative filtering algorithm, which shows each person a relevant eBook based on the behavior of similar visitors. The recommendation is boosted by each person’s preferred industry or function area (the use cases they have shown the most interest in while engaging with the site), excluding resources they have recently downloaded. For more information about how to build an algorithm, check out this eBook. It explains how to further refine base algorithms with boosters, as we did in this case.
Recommendations on our resources page function very similarly. But instead of just one recommendation, we have a carousel of items for each content type (e.g. eBooks, data sheets, webinars, etc.). The algorithm used here is quite similar to what’s used on the homepage, but it combines a trending algorithm with the collaborative filtering algorithm. The benefit of this is that if we don’t know anything about a visitor yet, we recommend the most popular content from the past 14 days.
So, for example, a visitor to our site that has shown a slight preference for demand generation uses of Evergage and is at the beginning stages of the buying journey would be shown a few related high-level data sheets:
Meanwhile, a visitor who has shown a very strong preference for retail content and is at the bottom of the funnel would be recommended data sheets for Evergage products, features and implementation services, focused primarily on those that other retail visitors have viewed:
We’ve seen some great results already from our newly implemented machine-learning recommendations. For example, by displaying more relevant content on our resources page, we’ve observed a 44% lift in content conversions! Our homepage recommendations have also yielded positive results; lifting content conversions by 40%!
SmartSearch gives our visitors the ability to find exactly what they’re looking for, typically after typing only a few letters into the search bar. The results are displayed directly into the search bar area. SmartSearch on the Evergage site has been designed to look at the metadata and resource/blog article titles while also leveraging the same algorithms as our content recommendations on the homepage (collaborative filtering boosted with industry and function affinities).
SmartSearch has proven to be an effective way to give customers a high-level view of our many and various resources and blog articles in a way that is relevant to each person’s affinities and/or industry. The image below shows a user experience of someone who just started exploring applications of Evergage for companies in the technology industry. As you can see, they are shown a wide array of resources (webinars, case studies and data sheets) that are relevant to someone just starting to discover how Evergage can be leveraged by technology firms.
Tips and Tricks for Demand Gen Marketers
We’re excited to have machine learning personalization on our website, and our results show that machine learning is not just for e-commerce. If you are a demand gen marketer looking to get started with machine learning on your own site, this post has hopefully offered some tips and ideas based on our successful experience.
As advanced and predictive as our machine learning recommendations are, there is still the need for a human touch. I found the best way to configure our machine-learning “recipes” involved some tweaking and a bit of trial and error, similar to cooking. This testing involved setting our recipes and then testing them for each industry and function to see if the recommendations delivered were significantly relevant. After some tinkering and tweaking with various recipes, we were able to dial in our recommendations for each individual user experience. Don’t be afraid to experiment and try new algorithms and combine them with different filters and boosters, as we did, for the best results.
To learn more about how Evergage can help you leverage machine learning in your demand generation efforts, request a demo today.