What are Recommendation Algorithms?
Recommendation Algorithms – Algorithms that help machines suggest a choice based on their commonality with historical data.
you decide to conduct a study on consumer behavior in shopping and take a survey of “people who ‘do not’ enjoy shopping”, there will only a meager percentage of them in the category; however, you take a headcount of “people who do not like to shop alone” and yes, your poll just changes drastically. Anyone who wants to shop, never ever wants to do it alone. This behavior of having “company” for shopping may on the outside just seem to be a characteristic of man as a social animal, but there is more to it than just that.
Shopping Experiences Today
Growing up we have always looked for the company for shopping. Just take shopping for clothes, for example, we have always asked for advice be it your siblings as kids or your besties at college or colleagues at work. Shopping trips traditionally were hours or even day-long trips researching the latest fashion, driving the bargains across various shops, and the try-out sessions. However, as time progressed, shopping trips started becoming a short affair and the besties and friends were replaced by the more “professional” personal shopper — who in turn could give you good recommendations for “the look” and “the image — the corporate meeting look, the cocktail party et al. Times were changing… are changing but one thing was sure, you still wanted the recommendations.
Currently, shopping trips have become even shorter and it just takes a few minutes and a few clicks on the internet. The recommendation and advice are coming as messages in emails or advertisements — exclusively tailored and personalized for you. The handbag that I shopped for the other day was less than a five-minute shopping trip — online. Generally, it takes me hours to choose a bag. This time I received some rather tempting recommendations of bags in my email and all I had to do was click and pay and wait for the delivery to happen. The catch here was that I had bought my earlier bags online and they knew exactly what I liked and did not.
What is a Recommendation Engine?
Recommendation Engines (also called as Recommender Systems) started off becoming popular in the retail industry, mainly in online retail/e-commerce for personalized product recommendations. One most common usage is for Amazon’s section on “Customer who bought this item also bought …”. Recommendation Engine is seen as an intelligent and sophisticated salesman who knows the customer’s taste, style and thus can make more intelligent decisions about what recommendations would benefit the customer most thus increasing the possibility of a conversion. Though it started off in e-commerce, it is now gaining popularity in other sectors, especially in Media. Some of the examples are YouTube “Recommended Videos” or Netflix “Other Movies You May Enjoy”. Other industries are beginning to use recommendation engines, such as the transportation industry. Waze uses it for intelligent navigation systems; IBM uses it for traffic control systems. Lately, GE started a Kaggle competition to find the best routes to save energy for the airline industry.
Recommendation Engine — Examples
Some examples of recommendation engine usage are seen in the following
- Facebook — “People You May Know”
- Netflix — “Other Movies You May Enjoy”
- LinkedIn — “Jobs You May Be Interested In”
- Amazon — “Customer who bought this item also bought …”
- Google — “Visually Similar Images”
- YouTube — “Recommended Videos”
- Waze — “Best Route”
Recommendation Engines as Filtering Systems
As we move into an era of data explosion, it is becoming more and more relevant to find ways to scan through the huge amount of data. Recommendation Engines become a great tool for filtering and ensure that the consumer gets to see the data that is relevant for his taste, his style and preferences and ensures he spends minimum time searching for the right data.
E-commerce / online stores carry a large product listing. If you want to buy an item on Amazon, you will find the listing in thousands, not just a few hundreds. Out of this vast sea of products we want to ensure that we present the most appropriate and the most relevant recommendation to the customer.
For a recommendation system to be good another important characteristic is it should be able to continuously learn and adapt itself flexibly to new user behavior. It also needs to be providing data real-time. For example, many special offers, changes in the assortments and price changes that happen make good recommendations obsolete shortly after having been made. A good recommendation engine must, therefore, be able to act in a very dynamic environment.
How do They Work?
Recommendation systems are based on algorithms that “learn” from past data. The data used may be about the products preferred liked or bought by the customer in the past or it could be products preferred, liked, or bought by “similar” customers. Based on this criterion the following types of recommendation engines are built.
- Collaborative Filtering
This is based on customer’s behaviors, activities or preferences and predicting what customers will like based on their similarity to others
- Content-Based Filtering
This is based on items liked by the customer and the keywords used to describe the items. It also takes into consideration the preferences chosen by the customer
- Hybrid Recommendation Systems
These are becoming popular where the combination of both the methods listed above is used. There is a trade-off that needs to be made in what to filter.
« Back to Glossary Index