Why does Netflix keep recommending movies I would be
interested in? Sometimes Netflix gets it wrong but other times it is spot on!
Well, after reviewing the reading, I decided to read some
basic articles about crowdsourcing. It was interesting to find that Netflix.com
uses crowdsourcing and has been using this type of resource for many years. It
has used crowdsourcing and variations thereof to understand what users want.
Netflix has harnessed the collaborative power of crowdsourcing and has
continued to provide a very human component in its computerized function.
It employs around 40 people to tag videos with at least 100
tags to work in combination with the recommendation function. Netflix also
tests different versions of its program with more than 10,000 users. If it
works with large crowds, most of the time it is then made available to all
Netflix users. In addition to these outlets, Netflix also has enlisted crowd laborers
around the globe to help caption online streaming videos.
Now it makes sense why Netflix ‘knows’ me so well when it
provides recommendations for me. I never really understood the effort behind a
simple list of recommendations but now I can see how extensive this process
really is.
It should be noted that Netflix was not always successful
when using crowdsourcing. In 2009, it began a competition to find the best collaborative
filtering algorithm to predict user ratings for films. The winning team
would receive 1 Million dollars. It turns out that after the team was awarded
the prize money, Netflix ultimately did not implement the algorithm stating “additional
accuracy gains that we measured did not seem to justify the engineering effort
needed to bring them into a production environment.” Netflix also experienced a
shift in its user base where more users began to stream videos rather than rent
them. As a result, the algorithm based on renting videos would not apply to the
user base that wanted to stream movies.
Personally speaking, I do see how my
preferences change when renting a video when compared to streaming a video
online.
In the end, crowdsourcing did not work for Netflix in that instance but that
does not mean that there is no true value with this technology.
Now I know why Netflix ‘knows’ me so well.
Great insight!! I always wondered that too and assumed they were tracking me according to my recently watched but I didn't know there was 40 people employed to simple tag movies and shows for that purpose. Very interesting!
ReplyDeleteIt is interesting that there are people out there tasked to just tag media. That kind of job probably did not exist not so long ago. What really is astounding is how they are able to mix that human component along with the computerized function and create something that works consistently.
DeleteWow, I had no idea either. What is going on behind the scenes?! And obviously there's money in it, right? They must see an increase in sales to justify their work. So it is freaky...or is it saving time and effort for everyone because then the consumer more easily finds what they want? Continuing food for thought....
ReplyDeleteIt is a little freaky, I would agree with that thought. I am not sure I want Netflix getting to know me too well. Yet, it does allow me to find what I want more easily with less time searching for similar movies.
DeleteIt makes me happy that they go through all that effort. I have considered Netflix as a standard and then wonder why Overdrive that I use for checking out library audio books is so off-base. Now I understand why. Overdrive is likely algorithm based on the last check out. Since the money isn't there for the people element, it just doesn't have the power of Netflix.
ReplyDelete