E-commerce retailers saw a tremendous gain in traction during the pandemic and now face headwinds as sales decline. In times like this, retailers need to go the extra mile to drive incremental online sales and to retain their customers in the long run.
Providing a personalized online shopping experience for shoppers is a great way of achieving this goal. Top e-commerce retailers like Amazon and Walmart are investing in building features that provide tailor-made experience for users, and these sorts of features can also be put to use by e-commerce retailers of all sizes.
Here are four ways that an e-commerce website can use long-term and short-term signals to provide a personalized shopping experience for their users. To be effective at personalization, it is also important for websites to have a robust way to track and store user activity.
Improve the Relevance of Search Results
The order in which the products are displayed in the search results plays a key role in determining whether a user is likely to convert on the query. Long-term historical signals such as clicks and conversions help the search algorithms understand the most popular products for a given search query, and what to display. Big players like Amazon have large data science teams that build powerful machine learning models that use these signals to improve the relevance of search results. However, smaller online retailers can achieve a great improvement through simple heuristic models that aggregate the number of times a product is converted for a given search query and display the products in the descending order of conversions.
Going a step further, building a crude user taste profile based on the attributes of products that the user purchased or clicked on is very powerful. For example, this can reveal that a user is vegan (or has a strong preference) if they buy a certain brand of products. This taste profile can then be used to improve the ordering of the search result to boost products that a user has preference for.
Present a Personalized Website to the User
The UX of many e-commerce websites are designed for users to explore and discover products, instead of directly searching for products. For example, many apparel websites have a homepage for users to navigate to various departments, aisles, or special collections like “Trending Now.” UX elements like dropdown navigation menus, filters for narrowing down options, high quality images to create the right visual appeal, and item detail pages play a key role in ensuring a seamless shopping experience.
The past browsing and purchase history combined with the user-taste profile can help to dynamically change the layout of the website and present it in a personalized way to the user. For example, e-commerce retailers can prominently highlight the most relevant departments based on the gender of the user and products they have previously clicked on. Similarly, they can easily implement a carousel similar to Amazon’s “Pick up where you left off” based on products that a user clicked or viewed but did not buy.
E-commerce websites can also provide social proofon product pages such as reviews, number of people viewing the product or purchased it in the last week, badges indicating that a product is a “Top Seller” in a given category, and more. You can extend this idea to generate personalized sign posts that grab the user’s attention. For example, show a “Best Deal“ badge to a user who previously purchased or browsed products with deals.
Or, create personalized collections of products based on the user's history similar to the “Deals inspired by your recent history” by Amazon. For example, if a user searches for “paper plates'' in their current shopping session, recommend a collection of disposable dinnerware products to the user, increasing the likelihood of users buying products that they have previously not thought of buying.
Big e-commerce retailers employ machine learning models which can infer these tasks and create highly personalized collections. Smaller retailers can partly achieve the same goal through simpler methods — by looking at the short-term historical signals like products that a user clicked on in the current session and recommend a collection of products that are in the same department to closely match attributes in the user’s taste profile.
Deliver More Value for Advertisers
Big retailers like Walmart and Amazon feature sponsored content on their websites to drive a lot of additional revenue. E-commerce retailers can work with third-party solutions to enable this capability on their website too. Historical engagement signals like past searches and purchases can be effectively used by the advertisers and the e-commerce retailers to match brands and users in a highly targeted manner. This will increase the Return on Ads spend (ROAS) for the brands who are advertising, and result in a lot of additional revenue for the ecommerce platform — which is a great win-win.
Improve the Quality of Inventory Using Historical Search Data
Inventory demand forecasting is very important for maximizing sales and profit. Retailers need to successfully employ multiple techniques based on historical purchase data to understand the demand for products. Analyzing trends in user’s searches is a great way of understanding what customers are looking for and not finding in your store. Those e-commerce retailers that deliver a truly personalized shopping experience will drive incremental online sales and retain their customers; they will be the big winners in 2023.
— Tejaswi Tenneti, Director of Machine Learning, Instacart