Chris Morton: How To Understand Your Customers
Personalized service isn't a new concept, but will continue to evolve as we adopt better techniques.
Five years ago, I lived in a house with three women, all of whom were addicted to online shopping. Curiously, they all shopped at the same sites, despite having radically different styles. It worked because these sites had a broad range of products there was something for everyone but unfortunately that made it hard for each of them to find the items they loved.
As a result of watching their frustration, my partner and I set up Lyst, a site that creates personalized shopping experiences for our each of our users, drawing from an aggregated inventory of the best fashion online. It’s based on the premise that shopping is better when it’s made just for you.
Personalization is at the core of what we do at Lyst, but it’s part of a wider macro-trend. Across the board, industries are starting to cater for the individual rather than the group. Medicine, which used to view people as more or less the same, now prescribes drugs on the basis of efficacy within different ethnicities, and is working towards understanding the individual’s genetic makeup to prescribe the optimal treatment. Many of us also consume information in a personalized way, using platforms like Twitter, which gives the user total control over which news sources enter their feed. In both these cases, it is technology that enables these personalized services – it makes things scalable where humans cannot.
Local commerce has been personal for years – shoppers have always been able to walk into a boutique and get a personalized service from an assistant or a personal shopper. However, it’s not the most efficient process. Consumers still have to explain their preferences, and online it can be tough to scale a human-powered service if you have millions of customers on your site every month. So an increasing number of retailers, marketplaces, and startups are turning to technology for a solution.
Here are some factors they should take into consideration:
Knowing your customer
Technology-driven personalization faces a litany of challenges. First, how do you find out about your customer? You can take a social approach and ask her to follow what she’s into, but that assumes she has to the time to tell you (on the internet, because no one has any time) and it also assumes that she is capable of expressing it herself – anyone who’s tried knows that describing your style isn’t easy.
Alternatively, you can study her actions on your site and algorithmically deduce her interests that way. Then, like Amazon, you can place her in a group of similar shoppers and suggest products to her than people like her also bought (known as ‘collaborative filtering’) although it takes time to harvest enough information to deliver a great experience, and the system breaks down when you buy a gift for someone else.
Or, like TripAdvisor, you can pull in information from third party sites, like Facebook or Twitter, but then you’re abstracting desires from conversations and social engagement – put simply, it’s not always easy to deduce someone’s style from their social media presence.
In reality, it’s not a choice of one over another. At Lyst, we use all three when we create personalized shopping experiences for our customers. Over-reliance on one technique can be limiting, while the best results — those that really help customers discover things they love — come from a blend.
‘Emotional commerce’ vs. ‘commodity commerce’
The blend depends on your industry. If your business is in ‘commodity commerce’ where the mantra is ‘faster, cheaper, better,’ range and decisions are made by the rational part of the brain, then the kind of algorithmic approach championed by Amazon and Netflix will probably work better. But if you’re in the ‘emotional commerce’ business, where selling is a seduction, and brands need to craft stories that resonate with consumers, then you should add more ‘social’ to your personalization mix.
In fashion, which is the space we focus on at Lyst, trends evolve irrationally – you simply can’t reply on past performance to dictate future desire. The shopper who always loved black, white, and navy may fall for a neon trend because a blogger she loves is championing it. An algorithm would have failed to show her the neon items that she now covets, but a social approach, where the shopper follows that blogger, delivers a more effective experience.
‘Emotional commerce’ verticals are also often dependent on influencers. Artists need validation from galleries and top collectors. This influence can propagate quickly online when consumers follow the influencers they like. For years, content sites like Twitter and Pinterest have been delivering personalized browsing experiences like this using social; now we’re seeing it more in the retail space, not only at Lyst for fashion, but also at Artsy for art and Etsy for crafts.
Another benefit of a social approach to personalization is that you don’t filter out serendipity as much. It’s tough for an algorithm to help you discover a new brand, but socially these sort of discoveries happen all the time when influencers share their newfound loves online.
Social graph vs. interest graph
Within social personalization, there are a couple of approaches – the social graph and the interest graph. Put simply, the social graph is your group of friends, such as who you’re connected to on Facebook. Conversely, the interest graph are the brands and people who influence you – some of them may be your friends, but the majority probably aren’t.
The challenge with deriving shopping recommendations from our social graph is that we typically only look to a small portion of our social graph for certain things. For example, I only want movie recommendations from maybe 5% of my friends – I don’t really care what the other 95% have to say about it. But I’ll turn to a different selection of friends if I want to go to a gig, or eat out. And it would be a nightmare if I got fashion recommendations from 100% of my social graph my friends each have their own preferences, and only some overlap with mine. In other words, to many people style is a deeply individual statement where they want to express who they are, and not be like everyone else.
The interest graph, however, is much better at delivering recommendations, providing it’s focused on a single vertical, like music or art. If I follow the mix of brands and people that influence me in each of those areas, I’m likely to get a better level of personalization.
Once you’ve found the right blend of social and algorithmic techniques for your personalized service, and sifted through all the data that’s available to you in those areas, the next challenge is where to deliver that service. The most obvious place is your website and mobile app, but you needn’t limit yourself. If you have a brick and mortar component to your business, the mobile handset is the glue that binds your digital and physical operations together.
Both Apple and PayPal recently announced low-cost ‘beacons’ – these are devices that sit in-store, with the ability to deliver messages to shoppers’ phones. This means that if you know what your consumer likes, you can now offer her personalized recommendations the moment she steps inside. Maybe some influencers she follows have suggested some sweaters on the first floor, or maybe an algorithm thinks she’ll love a new book by an author she read last year. The Contextual Support trend from PSFK’s Future of Retail report reviews the implication of products like Apple and Paypal’s beacons.
A thornier question is the idea of sharing consumers’ preferences with your competitors in order to build an even richer dataset for personalization. It’s hard to see it happen in practice, but the benefits for the cooperating retailers and the consumer could be significant.
Just ask Jeff
Jeff Bezos, who as Amazon’s founder is one of the fathers of personalized retail, ended his first annual report after his IPO by saying “it’s still day one for e-commerce.” He still closes his annual reports with that same phrase – and if it’s true for e-commerce, it’s certainly true for personalization.