If you are a sales associate in a retail store, you want to help each shopper find what he’s looking for so he can ultimately make a purchase. In most cases, you would have a conversation with him to identify how you could help. But what if you couldn’t speak to him first? What if you could only uncover his interests and his reason for visiting the store by observing him? In that case, you would need to pay close attention to how he interacts with every item in the store. In other words, you would need to track his behavior and draw conclusions about what that behavior means. Then you could use that information to determine how you can best help him.
That’s exactly what marketers do online when they track someone’s digital behaviors and use machine-learning algorithms to determine the best experience to display for him. But what behaviors do you need to track to help those algorithms make the best decision? There are several different types of behavioral data that can be layered together to provide a clear picture of an individual’s preferences and in-the-moment intent. In this blog post, I’ll explain the different types of first-party behaviors you can track and use for personalization across your different channels. First-party behavior refers to any action taken on your website, in your app or in response to your push notifications or email campaigns. It can be broken down into four major categories:
1. Site-Wide (and App-Wide) Behavior
Site-wide behavior encompasses general behavioral analytics about a person such as:
- Total number of site visits
- Total number of logins
- Total number of pages or screens viewed
- Total time spent on-site or in-app
- Average time spent per page
- Time elapsed since last site visit
- Total number of articles read, purchases made, videos viewed, etc.
While a bit high-level, this information can be useful in a number of ways. For example, you can apply it to target messages, notifications or experiences to first-time visitors or users. Here, the number of times a person has visited the site or app is important in determining the experience he receives. You can also use site- and app-wide behavior to target a campaign to returning visitors who have not engaged with the channel in a defined period of time.
2. Page Visit Behavior
Page visit behavior refers to data about specific page or screen views for an individual, such as:
- Specific pages or screens viewed
- Number of times each page/screen was viewed
While site-wide and app-wide behavior refers to such metrics as the total number of pages or screens viewed — or the average amount of time spent per page — page visit behavior refers to the specific pages visited or screens viewed. For example, page visit behavior would tell you that a person viewed Product Page A three times and Product Page B once. While still high-level, this data can be valuable.
3. Deep Behavior and Context
Deep behavior goes a step further than the previous categories and involves information about the page viewed as well as specific behaviors taken or not taken on that page. The information includes:
- Time spent on a page
- Mouse movement
- In-page context (category, tags, brand, color, keywords, etc.)
Deep behavior and context combines the level of engagement for each page with the attributes of the page itself (e.g., category, class, style, topic, keywords, etc.) to provide an accurate indication of an individual’s affinities, interests and intent. This type of information is critical for true one-to-one personalization.
4. Campaign Engagement
Finally, campaign engagement refers to the actions an individual has taken in response to any of your campaigns across channels. It includes behaviors such as:
- Personalized experience views and clickthroughs
- Email opens and/or clicks
- Push notification dismissals or clickthroughs
- Correlation of the above to device, time of day, or other variables
This data is immensely valuable, as it builds your understanding of each visitor and further informs your one-to-one personalization efforts.
Bringing It All Together
Let’s explore how these four behavioral data types differ with an example. Let’s say a customer visits a retail site. He’s a regular shopper and has viewed several product pages over the past month.
If you had access only to his site-wide behavior, you would only know such information as how many times he had visited the site, when his last visit occurred, how long he usually spends on the site and how many pages he has visited. With this information, you don’t know anything about his individual preferences or his intent for this site visit, but you could include him in a segment for loyal shoppers, for example.
If you had access to page visit behavior as well, you would know which specific pages he has visited. If his product detail page (PDP) visits were evenly spread across several different categories, brands and colors, you might conclude that he had no strong preferences in those areas. You could display product recommendations based on what other people also clicked on or purchased, but you could not truly individualize the recommendations to him.
But if you had access to deep behavioral data, you could understand much more. You could identify that he spent three times as long engaging with the shirt category than any other on the site. You could identify which brands and colors he prefers by how he engages with those product pages — and even which brands he does not like based on pages he views but immediately clicks away from. And you could identify when he has intent to purchase a product, such as if he spends several minutes engaging with a product page, scrolling through photos and reviews, even going so far as selecting a size — without yet adding the item to his cart.
Now, you could target individualized recommendations to him across the site or in email communications incorporating his specific preferences. And taking this a step further, if he does click through on those recommendations, adding one or two products to his cart or making a purchase, your campaign engagement data would automatically inform the personalization platform of this action. It could then, for example, remind him of items left in his cart and avoid suggesting items he already purchased when he visits the site again.
As the example above shows, while all forms of behavioral data are valuable, if you want to deliver true one-to-one experiences using sophisticated algorithms, you need deep behavioral data and contextual data — combined with campaign engagement data — to maximize relevancy.
While this intro to behavioral data for personalization is a great start, there is much more to learn about machine learning. Learn all there is to know about how machine learning has made the one-to-one dream a reality in our full-length book, One-to-One Personalization in the Age of Machine Learning.