Whether you're in-store or online, the benefits of a personalized customer experience are undeniable. We want salespeople who can help us find exactly what we're looking for as quickly as possible; we return to lunch spots where all we have to ask for is "the usual." For e-tailers, personalization has traditionally relied upon recommendation engines, algorithms that look to match prior consumer clicks and purchases with like-minded cohorts. And personalization is indeed effective for those e-commerce sites. According to a recent Infosys survey, 74% of sites who have implemented personalization technology reported an increase in sales, 58 percent noted an increase in traffic, and 55% observed an increase in customer loyalty.
In other words, a consumer journey tailored to a buyer's individual wants and preferences is a smart goal. But since current personalization techniques revolve around demographics and past buying and browsing patterns, they don't actually offer true personalization. Rather, they use the historical behavior of buyers to suggest products, both onsite and in the form of retargeted ads. This allows for a more personal experience, but not a true personal experience. To put it another way, it only buckets consumers into similar groups.
And while that approach has its benefits, it misses the mark in important ways. For one, recommendation algorithms don't serve new buyers without a purchase or browsing history very well. They also don't surface a large percentage of an e-tailer's catalog; by only showing products other users have purchased, these algorithms effectively ignore more appropriate products that may be exactly what the current buyer is looking for. Likewise, current personalization approaches don't account for unspoken and subtle preferences individual buyers have. Rather, they account for categories in a spreadsheet.
Many of these issues are born from the fact that, today, e-commerce sites (and e-commerce search) are designed around databases, not people. And while search and personalization have evolved in the past decade, they haven't evolved far enough. Recommendation algorithms can provide more personalized choices to shoppers, but they fundamentally do not treat buyers as individuals: they compare shoppers to each other.
Artificial intelligence (AI) solves for all these problems. It understands buyer intent in the moment, just like an in-store sales associate would, by interacting with the customers as they shop to more fully understand what they're really looking for. An AI-assisted shopping experience is more personal because it doesn't rely on similar (but different) users to recommend the right product. It looks at the individual shopper, in the moment, and helps them find the right item.
Why is that? Well, for one, deep learning AI can understand product characteristics at a granular level, even those that don't appear in a product database. Take a shopper looking for a red dress, for example. While current search and recommendation techniques require the buyer to search "red dress," then narrow that search by selecting a preferred brand, then re-narrow it by size and dress length and so on, until hopefully, the shopper finds a dress she likes enough to purchase. She's effectively browsing a spreadsheet.
With AI, that process gets both faster and more personal. That same buyer could simply click the image of a dress that appeals to her and AI will start making recommendations. Those recommendations aren't based on other user behavior but the actual characteristics of the dress she herself selected. That means that as the user keeps browsing, AI picks up on her preferences in real time. It can start making in-the-moment recommendations based on just a few choices without consulting a historical database that really is just comparing her with similar users. This allows for a much more personal experience that actually mirrors what an in-store experience is like. AI learns what she likes and it shows her dresses that are similar.
Since AI looks not just at product metadata but, importantly, at product images themselves, it can understand subtleties that haven't been tagged in the actual product database. In a typical experience, the buyer would have to select the characteristics of the dress she imagines in her mind's eye. She narrows her search by selecting preferred brands, dress lengths, sizes, and the like. AI eliminates those actions. Once a buyer clicks on that first item, it starts learning from each click what those preferences are. Because it looks at the image itself, it can understand preferences that every buyer has but might not even be able to articulate. As the buyer browses, AI can notice she's fond of a certain shade of red; in the database, all reds may be the same. AI can understand she's fond of a particular neckline, dresses with subtle patterns, ones with slender shoulder straps, and so on. Those subtle characteristics don't need to actually be tagged in a database. That’s because AI looks at the image files and identifies the most similar dresses hundreds of vectors and characteristics so it can make ever more accurate recommendations. This lets the buyer find not just a dress she likes, but the exact dress she wants.
What's more, AI-enabled browsing is just a more enjoyable experience. Customers just browse through an increasingly personalized catalog and, in a few clicks, they arrive at the item they really want. It's interactive, intuitive, and it allows them to find items they simply could not have otherwise found.
The best part is, this technology is not only available today, it's being implemented now. And you don't need an army of data scientists to enjoy the benefits. In fact, in most cases, all you need is a product catalog with some quality images.
In the end, it comes down to this: most e-tailers are looking to give their customers an increasingly personal experience. The data is in and that's just a good idea. But old methods can only get you so far. They group users as opposed to treating each buyer as a truly unique shopper. By using AI, it's possible to not only improve the actual experience of shopping on an e-commerce site, but help your buyers get what they really want. And at the end of the day, a delighted customer should be the goal anyway.
Andy Narayanan is vice president of Visual Intelligence at Sentient Technologies.