Recommendation Systems, explained easily.
How many times have you innocently looked up something on google, and immediately after…Boom! Instagram adds on it pop out creepily on your feed as if they were following your every move.
Some might call it tech magic, other businesses stalking us, but programmers know it’s actually a mixture of both, and its name: Recommendation Systems.
These filtering practices seek to find the perfect preference for you as a user, as they play an important role in finding products and content you care about.
They decide which post to display in your timeline and which new friends to recommend.
Ever met a salesperson in a shop constantly asking you about your size, model, style etc?
Well, think about RS as their digital siblings. They’re the driving forces behind platforms like Netflix, Spotify, Youtube, e-commerce businesses, and many more.
However, there are multiple approaches these systems use in order to make the ideal suggestions. Different studies and publications might group them uniquely, but here’s where they mostly boil down to:
Content-Based Filtering. – As the name suggests, it has information/content on both the items and user profile.
Generally, whenever a user creates his profile, the engine tries to get initial information about them. In the process, the engine compares the items that are already positively rated by the user with the items he didn’t rate and looks for similarities. Items similar to the positively rated ones will be recommended to the user.
Does it have any catches?
Yes, as it usually struggles with the phenomenon of a “cold start”. For a new user or item, there isn’t enough data to make accurate recommendations. The system needs to ask questions, conduct surveys, and obviously get to know the user. Kind of like being in the awkward stage of a first date.
Popularity Filtering. – It is actually the easiest way to build a suggestive system, by offering what is already popular. They can be the items that are mostly sold in an online shop or the reason you are destined to listen to trashy radio-hit songs. Still, popular nonetheless.
Collaborative Filtering. – Only concerned with “User behaviour”.
If two users have the same or almost equivalent rated items in common, then they have similar taste. Such users build a group or a so-called neighbourhood.
In this case, features of the items are not known.
Take Content-Based and Collaborative filtering and mix them thoroughly, perhaps with other approaches as well! This is what most systems use nowadays, as it also overcomes problems such as the cold start or sparsity problem. Teamwork truly does make the dream work.
The benefits of SR.
So why do businesses and social media platforms invest so much in them?
- They generate traffic and engage more customers.
- As these engines automate processes, they reduce the workload of your IT staff
- Personalisation and customer satisfaction.
- They provide relevant information and reports.
When someone feels like you understand their preferences and needs, they will be more prone to buy, watch, or consume the type of content you’re proposing to them.
Therefore you can now transform shoppers to loyal clients much easier.