What is Commerce Personalization?
How does it work? And where does personalization have the most impact?
For as long as I’ve worked in this business (and for many different commerce personalization platforms), this question remains a mystery to many. Even here at Sitecore, personalization is thought of as some kind of silver bullet technology that squeezes more revenue out of the same number of customers. Ask five different people and you’ll get five different answers.
It’s AI. It’s machine learning. It’s algorithms. It’s magic!
I’m here to tell you, it’s not magic and it’s not rocket science either. It’s data science to be sure, but it’s not difficult to understand. Let me prove it and explain what personalization is (and what it is not), how it works, and where personalization has the most impact. Let’s separate the industry lingo and marketing hype from the reality.
Commerce personalization refers to tailoring the shopping experience to individual customers based on their preferences, behaviors, and past interactions. It involves using data and analytics to customize the content, product recommendations, promotions, and overall user experience for each customer. By implementing commerce personalization, businesses aim to increase customer satisfaction, build loyalty, and boost sales by providing a more relevant and engaging shopping experience.
Personalization is a broad term that covers multiple technologies, but they can be summarized in three categories:
- Wisdom of the Crowd and Collaborative Filtering
- Individualization
- Segmentation
Wisdom of the Crowd and Collaborative Filtering
The “wisdom of the crowd” in commerce personalization refers to the practice of leveraging the collective insights and behaviors of a large group of users to enhance individual customer experiences and improve business outcomes. By harnessing the wisdom of the crowd, companies can offer a more engaging, efficient, and satisfying personalized shopping experience, ultimately driving customer loyalty and boosting sales.
Wisdom of the crowd can be applied to both Search and Recommendations. While Search can incorporate wisdome of the crowd data, it’s not best practice because it resorts the organic search results typically leading to lesser performance. Recommendations are where wisdom of the crowd shines!
Commerce personalization tools like Discover can provide simple unary product data sets based on aggregated product performance including Top Sellers by Items, Top Sellers by Orders, Top Viewed Products, and more. Although unary data sets provide recommendations for new visitors without history, these are not high performing because they display the same results for all customers. How personalized is that?
Binary data sets are a form of wisdom of the crowd and provide the product data for recommendations that are the highest performing widgets on most commerce sites. The reason for this is simple: Search is a shotgun recommendation approach based on minimal context (search keywords) that generates many results but few conversions. Recommendations are a rifle approach providing precise and focused products for display based on in-session click behavior and user history (clicks and purchases).
Illustrating the Power of Collaborative Filtering
The power of collaborative filtering is easy to illustrate. Let’s consider the example of a 16-slot recommendation widget that needs to be filled from a small product catalog of 150k items. What are the possible product combinations that collaborative filtering solves for?
That’s a lot of combinations (permutations)! And yet this is what collaborative filtering does: it provides the top 16 products in the best possible order for any one customer from all of these possible combinations. Now that’s powerful!
Individualization
Individualization takes wisdom of the crowd one step further. Instead of relying only on collaborative filtering algorithms, individualization further sorts product results based on individual customer product affinities. For example, if a customer is viewing a specific gender, category, brand or price range, personalization algorithms pick up on these patterns to influence Search and Recommendations.
Although this hyper-personalization appears to be a great benefit, it’s difficult to prove it’s value since most platforms have no ability to understand how often profile-based individualization fires. And since it requires a minimum of cross-session behavior to learn whether a shopping pattern exists, it’s also unknown how often it fires. Despite this lack of data, Discover best practice includes turning this feature on because tests indicate that it does no harm.
On the other hand, Search personalization can negatively impact performance since it resorts and changes the organic results generated by the search analyzers (algorithms). As a result, Discover best practice is to start out with minimal personalization and test into greater personalization. Some customers decide to turn Search personalization off because they don’t like how it impacts the customer experience; personalization can resort in-session search results making it difficult for customers to return to their prior search results and find previously displayed products.
Segmentation
Segmentation is also considered a form of personalization (which is why Sitecore’s segmentation solution is named “Personalize”). Although segmentation does not provide one-to-one personalized results in the same way Discover Search and Recommendations do, it provides additional personalization by viewing customers as part of predefined sub-groups based on customer behavior and demographics often defined by personas.