Website personalization faces several challenges, the foremost being privacy, a major consumer concern. On a nearly weekly basis, another bank, telecom, credit bureau or retailer announces they were hacked or had a data breach.
Most consumers acknowledge retailers must have access to some personal information, such as payment details and addresses—otherwise products couldn’t be purchased and shipped. However, they tend to be less comfortable with the idea of retailers possessing detailed data on shoppers’ likes, dislikes, clothing sizes and past purchases. In reality, retailers holding this type of information may be less risky than payment or address details, but the idea of organizations exploiting this data for marketing purposes can perceived as too personal, bordering on intrusive or even downright creepy.
Another major limitation lies in technology. Personalization depends on a coordinated, comprehensive understanding of the shopper, from their very first engagement through to user profiles, wish lists, store purchase data, email marketing and beyond. A single technology partner that can manage and automate these tasks doesn’t exist at the moment. Some of the most exciting personalization concepts will require coordination and integration between SaaS providers and shopping carts, and that isn’t yet happening with a high degree of efficiency.
Additionally, personalization algorithms are still young. At times, “personalizing” a user’s website visit can actually get in the way of their finding what they want. Since some tools work by surfacing relevant content and products, this means other products the user may actually be seeking, or be interested in, are virtually invisible.
Due mostly to privacy concerns, personalization hasn’t gotten very far at this point. Some social platforms have taken personalization a little too far, and in the process frustrated their most loyal users. Facebook uses personalization algorithms to customize individuals’ news feeds, and users complain that they’re seeing irrelevant posts, and missing photos from friends and friends.
If this creeps into ecommerce, serious usability issues will arise. For example, a shopper that searches for “blue sun dress” in the summer shouldn’t necessarily see blue sun dresses at the top of the dress category during their next visit in December. And an individual who bought a pair of Nike athletic shoes shouldn’t only see Nike products during their next shopping session. The fact that they purchased Nike during their previous session isn’t necessarily an indication that they prefer that brand.
What are some practical applications of personalization that are possible today and in the near future?
Historical product recommendations
Current algorithms simply display products that others have viewed or purchased after viewing the same product or page. This is effective and relevant much of the time, but not necessarily personal.
Amazon does things a bit differently, and their methods are much more personal in nature. Due to the fact that most Amazon sessions are by logged in users, Amazon can extract and leverage more personal information for use in later sessions. They know what the user searched for, viewed, saved and purchased during their previous visits. Amazon’s recommendation algorithms can understand what specific categories and products are most interesting and valuable to individual visitors.
To a degree, this limits the so called “creepiness factor” since the system doesn’t appear to have an uncanny ability to guess what the shopper is looking for next. Instead, it simply displays products the user has already shown interest in. Essentially, this is nothing more than page view history that is displayed on Amazon’s home page. While it’s not quite as easy, techniques are evolving to display recommendations like this for visitors that are not logged in.
Automated default sorting
Currently, all shoppers see the same default set of products when they browse a category or enter a search query. Generally speaking, the most relevant products are listed first, and users are given the option to sort and filter based on their needs and preferences.
In the future, it will be possible to guess what the best sorting and filtering options are based on demographic or historic data. Return visitors could be shown a different default sort. For example, if they have previously preferred to sort by low to high price a significant percentage of the time, this could be the default.
Again, care must be taken in order to avoid creating friction for the shopper. Demographic information such as age and gender won’t necessarily be useful. While the average 19-year-old female may prefer a certain cosmetics brand, elevating that brand over others may be obstructive.
Personalized search results combined with user managed profiles
Personalized search results based on past purchases and searches can be useful. But, this can also introduce unexpected usability challenges for the end user. Say a user prefers to purchase shoes in women’s size 8. It would usually be helpful to show only shoes available in that size, but if they’re shopping for a relative, this might hide shoes they want to see.
It’s also critical to make it obvious to the shopper that their results are personalized through a highly visible interface or window. If the search results are being tailored, there could be a message near the top of the page, or in the filter that tells them, and allows them to quickly toggle it off, or customize the level to which it’s being tailored. Providing the shopper with helpful controls can make the shopping experience more enjoyable and convenient, creating a stickiness that keeps them coming back.
Frequent online shoppers often use Pinterest to create collections of products they want to purchase. Some stores are already finding success by mimicking this experience within their store. Sites such as Kit.com are based entirely on sharing product collections. A rudimentary version of this would simply be to allow users to create wish lists. This can be expanded upon with controls that allow the user to save, share, and export their list. David’s Bridal, Target, and many others have implemented this in the form of a wedding registry, but there are other helpful applications of this concept.
Wish lists are one of the most basic, yet effective tools for providing a basis for personalization. Automatically, the site knows what the shopper is most interested in, through a tool that is perceived as helpful and additive – as opposed to intrusive.
Personalization in the form of demographic merchandising/targeting is a basic, non-intrusive yet highly effective form of personalization. Instead of analyzing past shopper behavior as a basis for targeted marketing and communications, shoppers are segmented simply on the basis of their known address details. The site can then tailor special offers, campaigns and marketing messages accordingly—for example, showing a winter sales banner only to customers in cooler regions.