‹ back to all posts
Guide to E-Commerce Personalization Systems

Guide to E-Commerce Personalization Systems

Personalization – a word that has become pervasive in retail marketing. The idea of personalization has become so ubiquitous that it has lost meaning, often used out of context and becoming its own white noise. Various personalization systems are designed with very specific goals in mind, whether that goal is suggesting products on site, generating happy birthday emails, or engaging a customer who abandoned their shopping cart. There are three basic types of personalization systems based on the goals they are designed to address and the mechanism through which they accomplish these goals. Understanding how they operate will allow you to make the best use of them and ensure that you are investing the right resources to meet your business needs.

Personalization matrix

Recommender systems: Recommender systems operate within the context of a specific transaction suggesting products on site when the user is browsing or as part of transactional emails. The most common use case is on the product pages: if you’re looking at an item, a recommender system may suggest other similar items in different styles or from different brands. Suggested products try to closely match customer’s current purchase intent (currently browsed product) naturally increasing conversion rate. The same concept is further extended offline to transactional email: 1) If you put an item in your shopping cart but don’t end up buying the item, you’ll get an email with recommended products similar to that item. 2) When you make a purchase, you get a purchase confirmation email with product suggestions that go well with your purchase. All of these use cases serve the purpose of increasing conversion rate as well as up-selling/cross-selling.

Contextualization systems: Like recommender systems, contextualization systems are designed to increase conversion online and operate in the context of an individual visit/transaction. Examples of contextualization include layout and creative A/B testing (Optimizely), navigation and product contextualization based on a search query (Bloomreach), and dynamic creative selection based on geography or visitor navigation path (for example, visitors from New York will see a banner with raincoats and visitors from California will see a banner with t-shirts). The idea behind contextualization is to choose the most appropriate landing page or layout to maximize conversion rate for the current visitor. While the goals of contextualization and recommender systems are very similar, they accomplish these goals through completely different means.

Customer lifecycle management: Whereas recommender and contextualization systems are designed for conversion and up-selling in the context of an individual visit or transaction, lifecycle management systems are designed to drive engagement across the entire customer lifetime (hence, increase customer lifetime value). They typically work through offline means like notifications and emails based on triggers specified via business rules. Probably the most common triggers are welcome emails sent out on new customer signups. Birthdays, customer inactivity, app downloads, and purchases are other examples of triggers that can result in lifecycle emails. Lifecycle management systems like Responsys usually do not tailor to individual transactions, but rather work to keep customers active throughout the year extending their lifetime value.

Putting it Together

Given the three fundamental ways to boosting e-commerce revenue, we can classify the three types of personalization systems into two groups based on the goals they’re designed to achieve:

  • convert / upsell (recommender and contextualization systems)
  • engage / increase customer lifetime value (customer lifecycle management)

Furthermore, we can then sub-divide individual use cases into online and offline: operating in real-time when you have user’s attention (e.g., on site) and trigger-based when you’re trying to get user’s attention. If we put it altogether, we get the following matrix:

Final result

Whereas the use cases in the first column are within the context of a specific visit/transaction and hence tailor to the customer’s specific purchase intent, the use cases in the second column have no explicit purchase intent from the customer and are designed to drive engagement.

If you pay attention, you notice that one of the four cells in the first chart is missing. Online personalization without explicit purchase intent. If the customer is online, shouldn’t the customer always have purchase intent? The answer is no. An example is an e-commerce newsletter delivered via email. While email has been traditionally viewed as an offline mechanism (trigger that pulls the user back online), Jetlore brings this experience online. With Jetlore, the merchandise inside of the email is personalized and targeted in real time as the user is viewing the email.

Such experience is different from traditional e-commerce recommender systems that tailor to user’s individual visit or transaction. In fact, using a recommender system to personalize items based on the customer’s last action on site will make newsletters repetitive and boring. We don’t have an item in the shopping cart; customer is not viewing a specific product; it’s not a “convert / upsell” scenario. Rather, the goal is to engage the customers, get them in the shopping mode. As a result, instead of personalizing to the customer’s individual action, we need to personalize to the magnitude of user activity and purchase history while paying attention to the real-time context like inventory trends, seasonal purchases, and “hot” items going on sale. 

In fact, online experiences with the goal of engagement are not limited to email: direct site visits landing customers on the homepage (22% of all online retail orders) and mobile app home screens have a similar challenge. There is no explicit purchase intent, yet you need to engage the customers to shop while you have their attention online. As we are entering the mobile commerce era, we believe majority of user interactions will fall into this scenario. There are actually hard facts behind this thesis outlined in one of our previous posts: the bulk of mobile shopping starts from either email (27%) or the app home screen (33%).

60% of mobile purchases do not start from an explicit purchase intent from the customer. Because of this, the next generation personalization systems, like Jetlore, will not need an explicit purchase intent from the customer but instead will make real-time predictions about what each customer might need from a comprehensive historical trail of their activity and real-time context.