I'm not so sure about that. Netflix has enough in-house experience in recommendation algorithms that they could pull something together without purchasing a 3rd party solution.
Plus you'll have two problems with Profiles. One, each user could have their own profile and therefore machine learning isn't going to detect abnormal viewing patterns. Two, some people choose to share a single Profile and so it'd be unpredictable whether you could deduce unusual viewing habits from that.
It's likely they will go based off of unique device type and location/IP. If you have a Smart TV streaming from an IP that's 3 states over, it's more likely to be sharing than if it's a phone.
> Plus you'll have two problems with Profiles. One, each user could have their own profile and therefore machine learning isn't going to detect abnormal viewing patterns. Two, some people choose to share a single Profile and so it'd be unpredictable whether you could deduce unusual viewing habits from that.
They'll just do it from account not profile.
And Netflix has external 3rd parties contributing to their recommendation teams and their geofiltering and proxy detection teams...
If you have 4 profiles watching wildly different topics, how would anything stand out enough to flag as another user? I think you've just added to my point.
Plus you'll have two problems with Profiles. One, each user could have their own profile and therefore machine learning isn't going to detect abnormal viewing patterns. Two, some people choose to share a single Profile and so it'd be unpredictable whether you could deduce unusual viewing habits from that.
It's likely they will go based off of unique device type and location/IP. If you have a Smart TV streaming from an IP that's 3 states over, it's more likely to be sharing than if it's a phone.