How Site Search Could Kill Amazon
Amazon.com is without question the e-commerce leader. Consider that Amazon was expected to capture nearly half of all U.S. e-commerce sales by the end of last year, reaching a staggering $258.2 billion, up nearly 30 percent from 2017, according to a survey from eMarketer. However, according to Eli Finkelshteyn, co-founder and CEO at Constructor, an artificial intelligence-first search-as-a-service provider, there's one area of Amazon's site experience that leaves it vulnerable to being dethroned: site search. Learn more in this interview that Total Retail conducted with Finkelshteyn:
Total Retail: What are the flaws that you see in Amazon’s on-site search?
Eli Finkelshteyn: Search is an area Amazon focused heavily on in its early days, going so far as to create A9, a separate company subsidiary focused on just this. Amazon knows how important search is, but lately its focus has shifted to other areas and it has let what was one of the best search experiences in the industry stagnate. Key developments like phonetic typo-tolerance and personalization within search have been ignored, so even slight misspellings like “kookies”’ leads to irrelevant results like dumplings and lotion. Internal studies at Constructor show on average 12 percent of searches have typos. At the same time, I buy the same brand of chips on Amazon every few months, but when I search “chips,” I still have to dig through pages of results before I find my brand. These are flaws Amazon’s competitors can capitalize on to give users a better experience, where they can be sure they will find what they want faster and will come back to shop again.
TR: Does search really impact the customer experience? What do companies have to lose with poor on-site search?
EF: We coined a term called “shadow churn.” What we mean by this are the myriads of people who come to an e-commerce website for the first or second time to give it a shot, don’t find what they’re looking for, and then never come back. Sometimes the company genuinely doesn’t sell the product the consumer is looking for, but other times they do sell the product. It’s just buried from a combination of a poor search engine and poor naming and product descriptions. Consider a time you wound up on an e-commerce website you don’t usually visit, did a product search and didn’t find what you were looking for. Did you ever visit that website again?
TR: How can other retailers and e-commerce sites step up their site search to zero in on Amazon?
EF: More ways than we have time for here. The big ones we recommend are as follows:
- Focus more on not just showing results that are relevant to a query, but showing results that will also be attractive to the user searching that query. Try searching for “laptop.” Amazon sells thousands of laptops and showing any of these would be relevant to the query. However, for most users, showing an overpriced laptop or one from 2013 that no one buys anymore would be less likely to result in a purchase. It’s important that the search results show laptops, but it’s more important that they show products people actually purchase.
- Conduct personalization that’s done algorithmically rather than via manual segmentation. Many companies now say they do personalization, but what they really mean is just that they will populate slightly different results for a manually defined segment such as “all people from Miami.” This is something to watch out for. Newer personalization, like what Google and Facebook use in their site search, learns from the preferences of each individual user and tailors results just to them. Watch out for any brand claiming it does personalization or AI where when you scratch the surface, you might find it really just does segmentation.
TR: What’s the future of site search look like?
EF: The future of commerce search will give relevant, personalized and attractive results to the user. This will happen to the degree that the user will have more confidence in the search functionality than they would asking the most knowledgeable employee at a brick-and-mortar store. Developments in natural language processing and machine learning are quickly bringing us to that point. The question that remains is whether companies will be able to build this search or partner with the right vendors before customers have one too many poor searches. The clock is ticking because users are always on the hunt for a better shopping experience and aren’t afraid to leave your site forever.