Have you ever looked up something online, and before you know it, you're being bombarded with ads on that topic everywhere you go? For example, let's say you're looking for the latest info on the next "Star Wars" movie. After watching the trailer, you start to see online ads for "Star Wars" T-shirts, "Star Wars" toys, "Star Wars" DVDs, "Star Wars" sheets... and a multitude of other "Star Wars" products that you never even imagined existed! This is all thanks to recommendation systems.

What is a Recommendation System?

Recommendation systems — also known as recommendation engines, recommender systems or simply RS — have been redefining the ways companies create customer experience. Recommendation systems have helped customers make informed and better purchasing decisions while making online purchases. If you have at any time made any online purchase, then you have almost certainly come across recommendations on products that are similar to the ones you have purchased. So, while you have been browsing products, the recommendation systems have been observing your browsing behavior and searching for products you might not already have discovered on your own. Recommendation systems play an important role in enhancing overall customer experience, especially in the online purchasing niche. Of course, it is good for business as well. Companies have been increasing their investments in improving their recommendation engines to help customers choose the best products.

How Does a Recommendation System Work?

Before we find out how recommendation systems have been impacting our lives, it is worth to know how they work and how they have been evolving.

A recommendation system is an information filtering technology, commonly used across e-commerce websites to offer filtered product choices to a customer who visits them. As the name suggests, the technology is used to offer recommendations on products that have similar characteristics. The objective differs depending on the perspective of the party. For a business that is selling products on an e-commerce website, it is improving its revenue-earning prospects by offering more product choices to a customer. For a customer, it is offering similar product recommendations and giving an opportunity to the customer to either buy a better product than the one already chosen or buy a product that can enhance the experience of the product already chosen for purchase. To provide the recommendations, the engines use a number of methods, including:

Collaborative Method

This method focuses on collecting information about the browsing behavior of the customer which includes, but is not limited to, products browsed, purchased, abandoned in the shopping cart and ratings. Based on this information, the engines search for products in the database that come close to fulfilling the customer’s requirements. The engines also analyze the interests of other users that are similar to the current user and may also recommend products other users have browsed or used. While this method is good to the extent of reasonably predicting the user’s choices, it suffers from the famous “cold start” problem, which means in the absence of substantial data, this method cannot work. The list of reputed brands that use this model includes Facebook, Twitter, LinkedIn, Amazon, Google News, Spotify and Last.fm.

Content-Based Filtering Method

This method focuses on collecting information about the attributes and characteristics of a product and then tries to find products whose attributes and characteristics are similar to that of the original product. While this method does not depend on user data, it tends to depend too much on products and does not focus on users. The list of reputed brands that use this model includes IMDB, Rotten Tomatoes and Pandora.

The comparative disadvantages of the above models have led some companies to use a hybrid method. Netflix is among the most visible brands that have employed a hybrid recommendation engine, investing more than $150 million every year.

Impact of Recommendation Systems in Our Lives — Case Studies

As stated earlier, recommendation systems have been impacting both the brands and their consumers enormously. To understand the impact, consider the following real-life cases of HealthTap, a startup focused on healthcare, and Airbnb, a website focused on vacation rentals.

Case Study: HealthTap

Problems HealthTap Wanted to Solve

Patients rely a lot on “doctor-recommended" medications. However, the doctor behind these recommendations could be someone who has been paid to provide recommendations of specific medications. In other words, it is a case of promoting specific medications regardless of whether they would provide the best treatment for the patient.

What Did HealthTap Do?

HealthTap launched RateRx, an initiative to provide medically qualified and independent drug ratings to patients. The app is available on smartphones and has more than 67,000 doctors in its network. These doctors review and provide ratings on drugs related to acne, anxiety, diabetes, headaches, arthritis and hypertension. When a patient views the details and ratings of a medication, RateRx can also provide recommendations of similar drugs with doctor ratings. According to HealthTap founder Ron Gutman, “It became evident that doctors were in a better position to share their educated and experienced views and recommendations on the efficacy and quality of medications based on years of experience prescribing medications and seeing their efficacy with a huge number of patients.”

Case Study: Airbnb

Problems Airbnb Wanted to Solve

Airbnb is a website where people can find accommodations for rent and also list their accommodations for rental purposes. According to Wikipedia, Airbnb has over 1,500,000 listings in 34,000 cities and 190 countries. Travelers are forever looking for cheap, comfortable and secure accommodations worldwide. Airbnb wanted to find ways to offer better, customized accommodation options to its customers. It wanted to know more about the unique requirements of travelers.

What Did Airbnb Do?

The main idea was to find out individual travel needs of travelers and give appropriate options or recommendations. So, Airbnb decided to dig deep into the customer data recorded in the form of travel reviews, accommodation feedback and other data recorded by the customers. Airbnb formed a team to do that. According to Mike Curtis, the vice president of engineering, “For a long time now, Airbnb has been an awesome place to go if you know where you’re going and you know when you’re going, but we realized that we have all of this data that other people don’t have. We have travel patterns. We have the reviews. We have the descriptions of the listings. We know a lot about neighborhoods that we can infer from the text in there.” So Airbnb got cracking with data and a recommendation system that gives personalized recommendations.

Evolution of Recommendation Systems

The hype around recommendation engines notwithstanding, they need to go a long way before truly capturing the user’s imagination. Right now, the engines follow a generic algorithm and do not quite offer tailored choices. The future lies in offering customized product choices to customers. For that, the algorithms need to factor in complexities such as sleep cycle, the user's mood, time of the day and energy output. It seems that the retail and the media industries are going to employ these engines the most and others will follow suit. The banking and financial industries, for example, are looking to increasingly predict their customers' next moves so that customized products could be offered. For that, a lot of data on things such as customer feedback, social media patterns, call center data, websites, emails and even education levels of consumers will be taken into account.

Conclusion

It will be interesting to watch how the future of recommendation engines takes shape. The algorithms that are used now have been in use for a long time, but businesses want more out of the concept. Brands are looking to tweak and improve their algorithms by constantly trying to make them more comprehensive. However, potentially the biggest challenge lies in the implementation of engines by industries that have not been traditionally using them, for example, the insurance sector which could offer insurance product recommendations.

Recommendation systems have the potential to help people in their daily lives in numerous ways, as well as help advertisers introduce products and services to wider audiences, and only time will tell exactly how this technology will continue to evolve.