The rise of online shopping, as we know it today, developed with the rise of the internet. As the internet improved, companies started to use it to give information about their products and services and advertise them. On 11th August 1994, with a Sting CD sold in the US for $12.48, the world’s first secure e-commerce transaction took place. Other products soon followed this practice, and online shopping has reached the current day. Through its evolution, one of the problems was that it lacked the human touch, human guidance, a.k.a. salespeople to help visitors through their store experience. Especially with the increased abundance of products provided on online stores, it has become a burden for visitors to go through catalogs or deal with irrelevant ads. Technology has provided websites with the best possible solution to this problem: intelligent recommendation engines, a system to help visitors through their journey on awareness, research, and decision making by, for example, reducing the abundance according to their needs.
A recommendation engine is basically a system suggesting the most relevant products, services, or information to the particular user in a given context by using data and algorithms. From Netflix’s “Other Movies You May Enjoy…” to Amazon’s “Customers who bought this item also bought…”, there are many websites and apps in the areas of music, movies, books, social tags, etc. using recommendation engines based on the behavior of the user upfront or behavior of similar users.
It’s no wonder that if set up and configured properly, recommendation engines bring benefits to both users and business owners. Users can enjoy more personalized products that meet their needs and get maximum customer support. And for business, since recommendation engines respond to users’ data, they drive more traffic by hitting the bull’s eye with the right suggestions and e-mails, by keeping shoppers engaged, and eventually by increasing conversion and sales.
A good recommendation engine should be flexible in responding to the needs of different websites/apps, scalable to keep up with increased traffic and work in real-time to reflect seasonal and regional changes and differences.
A recommendation engine is important because it helps businesses and websites to personalize their content, products, and services to the specific interests and needs of each user. By analyzing user data, such as browsing history, purchase history, and search queries, recommendation engines can provide targeted suggestions and recommendations that are more likely to be relevant and useful to the user. This can lead to higher customer satisfaction, increased engagement, and ultimately, higher conversion rates and revenue for the business. Additionally, recommendation engines can help users to discover new and interesting products and services that they might not have otherwise found, which can lead to a more enjoyable and fulfilling user experience.
One of the most important components of a recommendation engine is filtering models. According to Steven A. Cohen and Matthew W. Granade, a model is:
“A decision framework in which the logic is derived by algorithm from data, rather than explicitly programmed by a developer or implicitly conveyed via a person’s intuition. The output is a prediction on which a decision can be made. Once created, a model can learn from its successes and failures with speed and sophistication that humans usually cannot match.”
These filtering models used by recommendation engines are:
The recommendation engine using the popularity filtering model suggests products based on the number of likes, views, ratings, and purchases it gets. The most popular products are shown to all the users.
The collaborative filtering model is based on gathering and analyzing the information on user behavior, activity, preference, and predicting what they would like or need based on the similarity between the users. Although, as the user base grows, the system may work slower, the upside of this model is that the engine doesn’t rely on machine analyzable content; hence it can recommend complex items without understanding them.
This model is based on the assumption that people will continue to have similar tastes of the past in the future too. The filtering model simply deducts that if person A likes 1,2,3 and if person B likes 2,3,4, then they have similar tastes, so A must like 4 and B must like 1.
The content filtering model is similar to the collaborative filtering model, yet the correlation is between items instead of users. What is important is that product owners should define a strong similarity frame. The algorithm takes less time than collaborative filtering since generally, items have a smaller base than users.
The items that appear together in the same session are recommended to other users. So, the idea is similar to the content filtering model, yet the focal point is sessions instead of items. This model’s downside is that it offers less personalized items to users than the previous models.
This model combines different filtering models to achieve better performance. One example would be combining the collaborative filtering model with content filtering through various approaches to achieve the desired result.
Recommendation engines use the above-mentioned models according to the needs of the business and the users. They help users avoid unnecessary information, hence the overload. However, business owners should keep in mind that the effectiveness of these systems depends on not only what they recommend but also how these offers are presented.
According to Nielsen Norman Group’s research, users want to see personalized items over generic items. Some websites/apps such as Sephora place “Just Arrived” products at the top of the page, yet since personalized content is more valued than other content, users made comments such as:
“Maybe closer to the top though, maybe right under Just Arrived. Because a lot of times I don’t always scroll all the way down the homepage, normally I’ll look at the first couple of things and then go to the section that I’m trying to buy something. There’s a lot of stuff on this page and the Recommended For You is the second-to-last thing. So, I’d probably miss it. This is actually the first time I’ve seen it.”
Thanks to industry leaders such as Amazon, users’ expectations of personalization in retail have grown. According to a survey of 1,000 US adults by Epsilon and GBH Insights found that the vast majority of respondents (80%) want personalization from retailers.
Separating the recommendations into categories helps users to find the most relevant item more easily. This is useful because one user might have very different interests and because the platform might have a wide variety of content/items. One user comment is this:
“I used to put a lot of effort into managing the books Amazon would recommend, and then too, I buy lots of different kinds of books and the task of trying to tell Amazon through thumbs up or thumbs down or even giving 5-star ratings, it just didn’t really work. Because I could 5-star a novel for being a really good novel, but then what do I do for the technical book that was really good for the purpose I needed it for but isn’t like, I want to read this for fun.”
For users who want to manage the recommendations, the platform should have that option to get feedback and to allow them to edit the data. Users might look for items that aren’t in their interest but planned as a gift, so these users wouldn’t like to be recommended with these items. One way to allow users to manage recommendations is to allow them to edit browsing data.
Allowing users to manage recommendations is most effective when the updates start working quickly since users would expect to see the results immediately, especially if they give negative feedback.
While prioritizing the items and content and categorizing them, websites and apps use different layouts such as grids, lists, carousels. After Snapchat’s introduction and other social media platforms’ adaptations, a visually appealing and interactive way that can be used to assist recommendation engines are stories.
In-app stories increase the user experience by helping the recommendation engine provide their offers in a more appealing, creative, and familiar way. Many users know exactly what to do when they come across one. Since stories take the whole screen, users see only the recommendation screen without any competition for attention.
Check out Storyly to discover the power of in-app stories and how you can use them to find your users’ cup of tea.
When it’s well-executed, recommendation engines can increase conversions and engagement and improve user experience. They are increasingly integrating into our lives and affecting decision-making processes. Hence, it is important to understand how they work and how the brands can benefit from it by putting the users upfront. Businesses should keep in mind that putting users upfront happens not only by having the best engine but also by delivering the output to the user in the best way through prioritizing the user experience.