Being the world’s repository of random knowledge, the internet has a reputation for stimulating serendipitous discovery – a reputation born at a time when search engines and content marketing were in their infancy. Today, given that so many retailers use product recommendation systems, there’s a fear that the serendipity is dead; there’s a fear that rather than encouraging people to explore, personalization pushes customers into narrow information silos.

For example, imagine a shopper who searches for a party shirt and also subscribes to the retailer’s newsletter. She subsequently receives an ongoing stream of recommendations for party shirts and ‘people who bought party shirts also bought ……’:-  her first browse behavior has left her in a self-perpetuating loop of partywear.

In fairness, that type of personalization is typical of much of what’s on the market today. Rudimentary personalization often depends heavily on simple text analysis and metadata, which gives only an approximate picture of customers tastes, or worse still, personalization based on a single datapoint.

Advanced Personalization

Advanced personalization, however, requires a more detailed understanding of the customer’s tastes and interests. For instance, a personalization engine that uses Natural Language Processing can infer a product’s attributes: for example, it can use ‘semantic distance’ to understand that Adidas is closer to Nike than it is to Timberland.

In the end, advanced personalization creates a nuanced picture of each customer so that it’s possible to show every customer a wide variety of products that are both fresh and relevant.  As an analogy, suppose you’ve watched and liked Sully, American Sniper, and Inferno, a system might infer that you enjoy blockbusters (which describes all three movies), films directed by Clint Eastwood (which describes two), and films starring Tom Hanks (also two). Based on your interest in films starring Tom Hanks, as well as your interest in dramas, Bridge of Spies would probably receive a high recommendation score, while Toy Story, which has less in common with the other films, would likely receive a lower score. This content methodology enhances the customer’s experience and feels more conversational and human.

Enter serendipity

A detailed understanding of each consumer sets the stage for retailers to provide customers with a feeling of serendipity. In general, serendipity can be defined as the act of finding something valuable or delightful when you are not looking for it. Using advanced personalization to generate a sense of serendipity is all about creating a sense of ‘surprise and delight’ for customers by enabling them to discover products that as if the products are selected by their own personal shopper, and seem to be almost preternaturally selected for their unique taste and style.

Beware fake Machine Learning – what it is and what it is not

As brands become increasingly alert to the advantages of machine learning, naturally many technology vendors are getting on to the bandwagon.  When looking into Machine Learning, here are some quick questions that will allow you to choose whether somebody offers true machine learning personalization capabilities:

  1. How will you capture and rank the lifetime tastes of every customer? Can the platform create a ranked word cloud based on the tastes and preferences of every unique customer? Including both real-time and historic data, including offline and online data. If yes, do you purely use meta tags or can you use natural language processing too?
  2. Is it possible to also rank categories for each customer? Our site has more than 100 categories in the catalog but we only want to display c. 5 in each newsletter – how would your system select which categories to show| each consumer in a newsletter?
  3. Is it possible to automatically identify trending products? Additionally, rank those trending products within the context of the consumer’s ranked tastes?
  4. We wish to keep our email content fresh. How would you employ diversity models within the framework of the consumer’s ranked tastes?
  5. Feedback loop: what data does the platform incorporate to continually update the personalization algorithm?
  6. We have a small but agile team – how will this impact campaign production times?

Building loyalty by creating serendipity

When there are so many products at the customer’s fingertips, creating that sense of serendipity goes a long way to building the customer’s loyalty. They are more likely to keep returning to your site if they view it as a place to discover new gems, rather than somewhere to shop when they have a specific product or purchase in mind.

Plus, giving customers that sense is the key to creating a more personal relationship with each customer.  Done right, personalization is the difference between being a stranger who makes informed guesses about the customer, and being the personal shopper who introduces her to products she knows she’ll like.

Which shopping experience would you rather have?

 

 

Read more: ‘The 7 deadly myths of AI marketings’ »