Marketing
Personalization - Identifying Anonymous Users through On-Site Behavior
June 04, 2018 Santiago Oleas

Personalization and Contextual Marketing is very likely on your radar. According to Gartner’s 2016 Marketing Technology Survey sixty percent of marketers surveyed reported that they were using or planned to make use of personalization within the following 18 months. If you have not yet taken inroads to adopt a strategy, you are at least aware that this is the differentiator that will very likely define who the market leaders are. The Amazon and Netflix stories are well documented, describing how they are able to sift through millions of articles or titles and present each visitor with a good match ensuring they remain both relevant and profitable.  Today’s advanced Content Management Systems (CMS) allow you to fully cater to your customers and visitors by allowing you to offer a personalized experience while they interact with your brands. How does this work and how can you achieve this?

 

In this article we will examen an example of how we identify anonymous visitors through on-site behavior. If we can observe what our visitors are interested in, we can deliver relevant content using personalization algorithms that are often used in Content Management Systems (CMS). Sitecore is a leader in this space and Coria has a strong practice built around it.  The goal is to help you better understand how this is done so that you can gain a clearer understanding of how your knowledge of your products and your market segments could be used to provide the best experience possible for your customers - who are themselves expecting personalization and are often turned off when you are not able to provide it.

 

In this example, we will assume that an e-commerce site sells clothing for both men and women.  The content is typically organized by product categories and these categories often serve as navigation menus so that visitors may easily find what they are looking for. This site also provides additional content in the form of blog posts. What we are looking to do with personalization is showcase the most relevant content and products to your visitors by identifying who they are and to develop a contextual understanding of the purpose of their visit.

 

Looking at the content and product examples here, we can identify several traits that would be relevant to our customers such as Gender and Dress Code. These are known as Profiles or Profile Dimensions in Sitecore. They represent a dimension that differentiates content by areas of interest or relevance as they pertain to visitors. Each Profile has several Profile Keys that are how we could further identify each Profile. For Gender the Profile Keys are “Male” and “Female” and for Dress Code the Profile Keys are “Casual” and “Formal”.  In each of these cases we will create marketing taxonomy for the products and the content will also be scored on a scale for relevance. We will use a scale from 0 to 10 and assign them to these.

 

In this e-commerce example there is a homepage that features multiple types of content for a purely anonymous and unknown visitor, a blog post pertaining to music festivals and product pages for a women’s skirt, a women’s bikini, a men’s suit and a men’s t-shirt.

 

Sitecore provides a way of creating Profile Cards, which allows the combination of Profiles and Profile Keys with preset values. For example, the men’s suit below has a combination of Profiles and Profile Keys that were assigned identifying the product as Male and Formal. If the e-commerce site has more than one suit it would be tedious to manually set these individually. A profile card could be set with these values and assigned to all suits and similar products. 

 

Home Page

Festival

Skirt

Bikini

Men's Suit

Men's T-Shirt

Customer segments need to also be identified. By identifying your key market segments, you likely have an understanding of who your customers are and how they differ. If you are in fashion you have personas that are identified by how fashion-conscious they are, their budget and geographic location amongst others. 

Persona Definition

 

If you are in travel, your customers are likely differentiated by their taste for adventure, by their age group and their domestic status. Sitecore allows Pattern Cards to be defined and identified using the same Profile Keys previously defined, which represent these personas.

 

To go along with our content samples, we will use two simplified personas as examples to make it easier to follow.  Casual Connie is a female who likes to dress casual. Dapper Dan is a male who prefers to wear suits. Using the same characteristics we previously assigned to the content, we assign an expected value for each that represents who these personas are. Casual Connie is a female so she gets full points for Female and zero points for Male, for example. Sitecore allows Pattern Cards to be defined and identified using the same Profile Keys previously defined, which represent these personas.

Casual Connie

Drapper Dan

We can draw conclusions based on supporting information that Dapper Dan would primarily be interested in the suit and possibly the t-shirt. Casual Connie would be interested in the bikini and possibly the skirt. How does a CMS draw the same conclusion? The CMS system will take advantage of both implicit and explicit information for each visitor to determine what content or products are the most relevant.

A visitor navigating the site will consume content providing insight into their interest. By categorizing and scoring the content by Gender and Dress Code  we can get a sense of that visitor’s intentions.  Assume that a visitor navigates through the site and visits the following pages. Each page has a score based on each of the traits identified:

 

Traits

 

By the time our visitor reached the Bikini page we have determined an AVERAGE score for Gender and Dress Code to be: 

Anonymous Visitor

Gender:
Male: (10 + 5 + 0) / 3 pages = 5.00
Female: (10 + 9 + 10) / 3 pages = 9.67

Dress Code:
Casual: (10 + 10 + 10) / 3 pages = 10
Formal: (10 + 0 + 0) / 3 pages = 3.33

 

 

 

 

The system can utilize the accumulated score the anonymous visitor has accumulated determining the persona that is the best match. The Pattern Card that is “closest” is the best fit for the individual. The Euclidean distance formula, which may also be known as Pythagoras' theorem, is used to calculate this. The formula to calculate the distance “d” using the four different variables that represent Male, Female, Casual and Formal:

 

The “p” values are from the Personas we have and the “q” values are from our anonymous visitor:

Values

 

CASUAL CONNIE =  Connie Value

 

Values 4

 

DRAPPER DAN = Dan Values

 

The anonymous visitor is determined to be “closest” to CASUAL CONNIE as a persona than to DAPPER DAN.  Looking at radar graphs that represent the Personas and the Anonymous Visitor a similar conclusion may be reached visually. The anonymous visitor’s accumulated interest and intent matches closest with CASUAL CONNIE than it does to DAPPER DAN.

 

Graphs

 

This provides us the ability to personalize content on the website that is more relevant to the individual. Using content that has been tagged with the persona CASUAL CONNIE, the system can serve up articles, blog posts and products that can engage the site visitor with relevant content. Relevant content gives visitors more reason to remain and to convert to goals we have set out for them.

The benefit of this over explicitly known information is that the most recent actions give us more context. Even if our visitor is logged in and is known to be typically a CASUAL CONNIE perhaps she is now shopping for her husband and is browsing for gift ideas. So even if we were to start with a score of Female and previous visits told us she is a CASUAL CONNIE, if we were to place more weight on recent activity we would be able to adapt to her current needs. If she is shopping for birthday ideas for her husband, we will be better off serving DAPPER DAN content if that is what she was browsing. Sitecore has concepts around decay, which allows us to place greater weight on the most recent activity by letting older visits and scores to have less weight than the most recent activity.

An important factor in all of this is determining how to categorize and score your content and how to build up your personas. Coria has a series of workshops available that help you make the best use of your CMS environments especially if you are using Sitecore. Contact us to learn more about these workshops and the services we offer.

Santiago Oleas
Solution Architect