Personalization is a major component of many online and offline websites and services. By understanding their users and producing content, services, or experiences that are more contextually relevant (yes, even ads), that are more enjoyable and engaging (such as a social network or watching Netflix) or that increase success (such as learning a language, finding a date, or getting fit), they increase their value to a user.
How a site demonstrates that that they understand their users, and how they help them , or help them to help themselves, will take many forms. Some will be true personalization, others, tailoring or customization. What’s the difference? This document sets out a framework to understand these various, related terms.
Wikipedia provides two senses of personalization. First,
the action of making something identifiable as belonging to a particular person, especially by marking it with their name or initials.
This simply means that a site addresses users by their name! We’ll call this name personalization. Communications should always be personalized in this sense: Dear John, Dear Abby… This is table stakes for any engaged and empathetic brand.
The second, more important, dictionary sense is
tailoring a service or a product to accommodate specific individuals, sometimes tied to groups or segments of individuals.
We’ll call this true personalization. The latter phrase “groups or segments” is key. There is a continuum of who different personalization efforts reach (see below). Some may only touch a broad group or persona, others will get down to very fine segments, ultimately segments that each contain a single user — this is the ultimate goal: a truly unique, bespoke, and contextually relevant and valuable experience for each and every user.
Let’s dig into true personalization versus some other related terms. We’ll briefly go through examples of a number of different types of personalization, wrapping up at the end with a set of decision criteria and a simple framework with which to compare them.
Imagine you are ordering a new car. You go to a website, are provided with a set of options and you choose the paint color, chrome wheels, and sunroof upgrade. The car is then made specifically for you. This is an example of customization. Similarly, Brian is getting too much email from his favorite e-commerce site. He updates his preferences to mostly opt-out, except for special promotions.
In customization, you make the choice, you are in charge. The site provides a set of options and then carries out your wishes, creating or configuring the item or service as specified.
Bob, an investment banker, asks his tailor to make a suit. The tailor measures Bob and make numerous decisions on his behalf, the length of the sleeve, the width of the lapel etc. and finally makes the suit.
Lucy, unfortunately, is in hospital recovering from an operation. Her meal plan is created specially for her tastes and her particular condition to aid recovery.
These are both examples of tailoring or custom tailoring. Here, you outsource the decisions to the expert. They decide what is best for you and create content especially for you.
Both customization and tailoring are instances of made-to-order.
Jane sees her doctor. After describing her symptoms, the doctor comes up with a shortlist of three drugs that might be appropriate and tells Jane which one she needs to take. A tax consultant determines how their client should itemize deductions. These are examples of expert advice.
Here, the expert makes the final decision, not the patient, and is more prescriptive, more of an order to follow, than tailoring. The list of candidate drugs was chosen specifically for Jane but the drugs themselves were not created specially, they were generic.
After chatting to a client traveler, a travel agent provides a set of day excursions that the user might like for their trip to Egypt. In the running store, the assistant suggests 3 pairs of shoes that the shopper might want to try, given their flat feet. These are examples of guided choice.
Here, the list of trips was chosen specifically by the expert but, like the drugs, the trips were not created specifically for the user. At the end of the day, however, the user decided which ones they wanted.
At the registration desk of a 5K run, a runner picks out an event t-shirt from three piles of different sizes. In the supermarket buying apples, Amy chooses Fuji over Gala apples. These are examples of simple personal choice.
The user made the decision themselves from a set of generic selections of static objects.
We’ve gone through name personalization, customization, tailoring, expert advice, guided search and personal choice. The final, and most important, category is true personalization.
Here, the service modifies the content or experience based on what it can ascertain from users’ explicit and implicit signals through an on-boarding process or their overall history of interactions with the service. Importantly, the overall program is dynamic — in contrast to the fixed customized-car, drug, or t-shirt. It responds to the user’s inputs and choices, optimizing the program for them.
When building a personalized experience, start simple and develop sophistication over time. For instance, Netflix’s initial recommenders were likely relatively simple collaborative filters and content-based recommenders that picked up on genre, actors, and some other basic attributes. However, now they are using additional approaches and data and they personalize the whole experience: the shelves of recommended shows, the billboard (large promoted show at top of page), and even the artwork of individuals shows. That progression took years.
We’ve detailed seven related terms. There are four key criteria that help us decide which is the right term:
- Decision maker: who decides? You or the expert?
- Bespoke choices: is the set of choices, i.e. the list itself, specific for you?
- Bespoke content: are the items associated with those choices bespoke (tailored suit) or generic (t-shirt)?
- Content: static or dynamic?
We can summarize the different terms in the following table:
Thanks to a number of colleagues in the WW (formerly Weight Watchers) data science team for their insightful comments and suggestions.