One of the original and dominant use cases for Big Data are recommendation engines.
The benefit of recommendation engines from a consumer’s perspective are two-fold:-
From an organisation’s perspective, recommendation engines assist at 3 points in the sales process
How do recommendation engines discover what you might like?
4 basic techniques are commonly employed before a product is advertised to a user:-
- Examine the contents of a user profile. For example, by looking at a person’s gender and age, you can filter out certain products which wouldn’t be appropriate.
- Examine what a user has already indicated that they might be interested in – Every time you like something on social media, a profile is built up about you.
- Combine a user’s preferences with those of their peers – Assumption is that if several people in a group like something, it’s likely that you will.
- Look at what a person’s bought in the past – If you keep watching the same genre of movies or the same actor, it’s good odds you’ll like an ad for a new film that matches that criteria.
Combinations of recommendation algorithms are also used to improve the odds.
How do you know your recommendation engine is successful?
When ad’s are placed on a website, the image is located on another website which counts each time the ad was presented. This is known as an impression.
If a user clicks on your ad (known as a click through), then they’re taken to your website and your website can detect their visit and capture information about the user’s ip address and referrer site.
Given this information, you can also work out how many unique visitors you’ve had in a particular period by simply de-duplicating the ip address information.
Your recommendation engine will be somewhat successful, if significantly more clickthroughs occur per 1000 impressions than with advertising campaigns not using the recommendation engine, and successful if more unique visitors actually make a purchase.