The last decade has seen new data-driven companies disrupting industries as diverse as transportation, retail and, of course, media. We-all-know-the-names, and the stories. Companies with vast amounts of data, and who use it to target their customers individually, has enabled these companies to offer personalized experiences to their customers - and dominate their respective verticals in the processes.
Lots of companies want to try and replicate this model. The starting point for many conversations our data science consulting team has with customers is that said customers want to move from relying on "traditional market research" to "big data" - or words to that effect.
We always applaud companies wanting to embrace data more, but we also counsel our customers against throwing the baby out with the bath water. Even in an era of big data, there is still tremendous value in traditional research methods. In this article, we look at how one company
Netflix and segmentation on steroids
In particular, we're always surprised at how quickly our clients sometimes want to move toward targeting each customer on an individual basis - "like Netflix" - instead of targeting them in groups. There are a few issues with this - it takes a long time and a lot of effort to go from one extreme - mass targeting - to the other - 100% personalised targeting - and is not a change that can happen overnight. But more to the point, even the most analytically advanced companies are still using audience segments to target their customers. Yes, even Netflix.
Netflix gathers vast amounts of data about the viewing habits of its customers which it then uses to help key decision-makers make the decisions on which programs and new seasons to invest in and which to reject. This data-driven approach gives consumers programs they want to watch, even when they consume the latest blockbuster season in a single night. What we find particularly interesting is how Netflix is using audience segments.
Netflix explicitly rejects using the traditional demographic segments typified by Nielsen et al. It makes no effort to segment its audience by age, gender, location, and so on. But it does have a sophisticated segmentation around content usage. Netflix has categorized users by their viewing habits into roughly 2,000 micro-segments, and it is this that partly powers the Netflix viewing algorithm. That's to say, what appears on your Netflix menu is not driven solely by you and what you like, but by your presence in these segments.
Netflix automatically categorizes any customer into some of these micro-segments. For example, why does Netflix need to know someone is in the Latino demographic if it can see by their viewing habits that this person mostly watches TV in Spanish. And, with those Latinos who prefer English-language TV, Netflix's micro-segments work even better. Consumer viewing habits draw up a more accurate picture of customer demand than Nielsen's segmentation by demographics, which arguably is what gives Netflix its competitive edge over its more traditional rivals. Netflix's segmentation also explains how, having clawed its way to the top through its innovative uses of data, it has managed to remain the most significant player in the OTT space.
Netflix's micro-segments are comparable to assigning an in interest in particular TV genres. Netflix uses what it describes as "verticals,f" which are very sub-specific genres. So while drama is a typical genre and both romance and period drama can be considered sub-genres of drama, a Netflix vertical would be period romance. They also have a vertical called young comedy though this does not mean that only young people will watch shows in this vertical, given that "young people" is demographic data.
These verticals have successfully allowed Netflix to capitalize on their initial growth, and remain tremendously popular. They have also used verticals in the structure of their organization. Genre heads - and sub-genre heads - within Netflix are widely understood to have more decision-making power than their equivalents at a conventional network. This way of working is what allows the company to produce roughly a thousand shows in 2018 alone, a particularly impressive statistic given it created its first show a mere seven years ago. Netflix also says that despite having many verticals shows which are difficult to define by a particular vertical, and therefore may have a broader appeal, are often the most successful new shows.
Lessons for the industry
Our data science consulting team likes to use the Netflix example with our customers to help them to understand the continued importance of traditional market research methods when it comes to understanding its customer base, and how to use TV data to target them.
What it demonstrates is that reaching the nirvana of personalized targeting does not happen without the creation of a lot of audience clusters or segments first-and all the data engineering that goes with that. It also reinforces a broader point that "old" and "new" methods of understanding your customers will continue to coexist for some time. In our industry, it also highlights the move away from demographics research, with companies instead using data and data science to create new and improved segmentation models, both for TV shows and the people who watch them. Or to put it another way, something old - segmentation - is a small step towards something newer - personalized TV - even though it may not immediately seem like it.