I don't usually like pre-fixes like 'hyper' as they rarely add much meaning (personalisation is just personalisation, right?) but I think the extent to which Netflix uses sophisticated content personalisation is sufficiently distinctive to justify the hyperbole.
In a recent post on the Netflix technology team blog they describe, for example, how they use data on individualised viewing history to not only personalise the rows of films and shows that are presented to you, the titles of the rows and the content that populates them (using tens of thousands of sub-genres), but how they also use static imagery from films to show different film cover images to different users in order to appeal to the myriad different tastes that they serve:
‘…we don’t have one product but over a 100 million different products with one for each of our members with personalized recommendations and personalized visuals.’
So the artwork chosen to represent a film to you may be influenced by anything from the kind of genres that you’ve shown an interest in previously to the number of times that you’ve streamed films with particular actors or actresses in them.
It's important that data like this is used responsibly of-course, but what's interesting about the way they talk about this is the challenges that they say this brings - not least that of making the algorithm that controls artwork personalisation work in tandem with the existing one that powers content recommendation. The former has to sit on top of the latter without disrupting its work. To work well, artwork personalisation requires a huge amount of data and a wide range of source material which can make attributing tangible value to algorithm changes really hard. An overly complex artwork personalisation system might impact the load performance of the home screen (Amazon use product combinations in their recommendation algorithms for a similar reason). And then the algorithm has to adapt as more content is viewed, making it an ever changing formula.
It’s a true engineering challenge but one that has brought measurable improvements to how users discover content. And I think there's an interesting thought here about the role for such sophisticated use of individualised content. The debate about highly targeted personalisation vs mass-market messaging in the media and advertising industry is often more polarised than it needs to be - we need both of-course. But we often suffer from the results of poor execution (like most retargeting), poorly aligned incentives and measures (it works well enough for advertisers or publishers not to be so concerned about trade-offs in user experience) and poor thinking (forgetting that if we don't like something, other people probably won't either).
Right now there are a surfeit of examples of poorly delivered personalisation and targeting. But that doesn't mean that personalisation itself is inherently bad. Alongside Netflix for example, I think Spotify's Discover Weekly has consistently given me considerable value over a number of years. So there are already some great examples out there of how sophisticated individualisation of content can enhance customer and user experience by no small degree, but given the availability of the data and the technology these examples are surprisingly few. Put simply, we need to better.