One of the common refrains I hear from both industry creatives as well as industry executives is that streaming metrics - How many people watch?, Who are they?, Why did they decide to watch? Was the release profitable? - are essentially unknowable things. It's a confusing time for people who grew up in an era where television was fairly easy to define. What were the ratings? What was the demo? Did the budget match the results?
One of the issues has been that the data streamers collect is rarely shared with their content partners or the public (including reporters). However, it has also been the case that the granular data necessary to provide streamers with more detailed assessments on the most accurate way to value content, reduce churn, and offer specific suggestions on where advertising will be most effective has been lacking.
One new option is SubScape Streaming Segments, a collaboration between Claritas and Magid which seeks to understand people’s streaming behaviors. The segments uncover whether people subscribe to streaming services for a specific show and churn once the season is finished, or stay subscribed. It also covers whether they’re a streamer's biggest evangelists or aren’t interested in streaming at all.
I recently spoke with Ron Cohen, SVP of Practice Leadership at Claritas LLC and Mike Bloxham, EVP of Global Media and Entertainment at Magid about their partnership and the possibilities presented by SubScape.
This conversation was incredibly helpful to me. But it also made me realize I could spend a year writing stories based just on the top-line data they've assembled.
This conversation has been lightly edited for clarity.
Mike, if I understand it correctly, the genesis for this partnership goes back to a couple of years ago when Magid started looking at data from people who were subscribing for less than six months to a streaming service.
Mike Bloxham: The company's been around for over 60 years now, doing all sorts of different kinds of stuff. One of the things we've been doing, along with everybody else for several years now, is trying to understand audiences as they relate to streaming platforms, what drives engagement, what inhibits engagement, and what drives growth.
Then, as churn became more of a thing, as growth slowed down, obviously we, like everybody else, started to look at churn. And there was a study that we put out, or that we did for ourselves, which looked at the whole video entertainment landscape, all of it.
And streaming became a greater and greater part of that. One of the questions we started to ask was not trying to quantify directly actual churn, but we were trying to understand why people churn. But one foundational question was asked of people who sign up to an SVOD.
We asked, when you sign up for an SVOD service, how long do you typically expect to remain a subscriber to that service? And it was really interesting because pretty much every time we asked this question, we would get an answer somewhere around 40% of people who answered with some period of time less than six months. And for some of them, we broke it down beyond that.
Their behavior was in and out, you binge a show and go. Others would say for a month or three months, six months, and beyond. That behavior represented a significant chunk of people and a lot of money.
So we decided we wanted to understand that better to see whether or not there was a there there. Is that group of people in any way meaningfully different from those who plan to stay longer than six months? We did a whole lot of analysis on the work that we collected, including matching it up with some other data sets.
And the short of it was that we ended up with a lot of different aspects of insight on this question. One of which was this segmentation, which became a very useful way to actually look at the entire population across all services, or all the services we were looking at, which was over 30 of them. And what we found was there was a lot to be understood about this group, they weren't one homogenous group, but they do exist. We all know that there are people who are habitual churners. 
They're much more likely than others to churn. Some people are serial churners, returners, whatever you want to call them. But we still typically tend to talk about churn in a very homogeneous fashion.
We saw that there are two very distinct groups within the streaming population of high churn subscribers, but they're actually very different. One of them is a segment of the population, which is a real growth driver. We call them hypers because they're hyperactive consumers of media, everything that they consume across the board.
They very much proactively curate their experience. There's a lot of FOMO driving their behavior. They hear about stuff.
They want to see stuff. They're looking for it. They're telling other people about it.
But they're the group that's learned it's really easy to turn it on and turn it off again, and they can always come back. It's not a big deal. There's another group who are also high churn, but they don't represent growth.
They're not hypers. They're high churn, high value. 
There's another group we call the Digitarians, who are high churn, low value. They tend to spend much more time on other digital platforms. TV's not as central to them as it is to other groups.
They tend to skew younger, unsurprisingly. But a lot of their time is on TikTok. It's on Instagram. It's on YouTube. It's Snapchat and all the rest of it. If you have certain types of content, they can be very useful, but they are statistically more likely to be jumping from free trial to free trial, using other people's passwords, or dipping in and out for something very specific.
They actually over-index, as we know from our data, but also from the Claritas data, they over-index for their interest in things like UFC and MMA. So if you've got that kind of content, well, that's a group you probably want to market to. But if you're pushing dramas and the conventional type, they're going to be less fertile for you.
You're going to get some of them, but you probably don't want to spend too much money getting them because they don't represent as much longer-term value. So that's what we started to see. There are obviously other segments that represent stability, as opposed to the volatility of the two that I've just been talking about.
But what it led us to was a very strong realization that you need to understand what proportion and who within your subscriber base represents stability, represents volatility, represents an inclination to tell other people about content more than others, because they're basically your net promoters. They're the people who define cultural relevance. They're the ones that say, go see this new show on Netflix or Apple or wherever it is, because it's fantastic.
They draw people in. So that's where we decided, we really need this to be actionable. A lot of segmentations are nice to have, and they give you great insights. But in order to be actionable, you've got to be able to buy those audiences. I was speaking at a conference and ran into Ron and some of his colleagues. We got talking and realized "these guys do exactly what we want to do with the data that we got."
How do we go about doing that? And that's where we really started working closely with Claritas and got on the path that we're now on.
Ron, I don't think people necessarily understand the level of granular data that you can put together. Can you talk about what some of that household data looks like? How deeply can you dive and provide a look into individual households and individual neighborhoods?
Ron Cohen: Sure. Claritas has an identity graph. It's got every household in the U.S., every consumer household in the U.S. It's got business data, too, but we'll leave that aside for now. And it's got data for all the adult individuals in every household. We have data about people who are not adults, too, but we don't report it. I mean, you'll count it at the household level. "Oh, this household has children." But we don't report them. We don't report data for minors, obviously.
But we know who's in the household. When I saw Mike's presentation, I think it was like two and a half years ago, it was probably one of the first times Magid presented SubScape. It clicked immediately for me.
I thought, we could link this to PRIZM. PRIZM Premier is our flagship segmentation system, if you're familiar with Claritas and PRIZM Premier, which exists now in our identity graph at a household level. So we have thousands of pieces of demographic information, literally thousands on every household, some of it's demographics, some of it's lifestyle, some of it's segmentation, some of it's neighborhood level, but lots and lots and lots of data, most of which goes into our models to make PRIZM Premier and other segmentation systems.
And I said, "If we can link that to what this guy is up there on the stage showing, Then we got something. Right after his presentation, I approached him. It really was kind of a match made in heaven. He's got the kind of research data that allows us to leverage our segmentation, and we've got the kind of data that lets his segments become actionable.
It takes their research and makes it actionable on the ground in the real world and not just theoretical, right?
This is a partnership really where it makes sense. You're not competing against each other.
Ron Cohen: Not at all. We're both leveraging the core strengths of the partner.
Mike, I want to talk to you a little bit about how you put together your data. Can you talk a little bit about how you collect that data, how you collate it and put it together in a way that is actionable for your clients? I know a lot of it's proprietary, but what are you doing to ensure your data is as accurate as possible?
Mike Bloxham: Our data is very rigorously put together. We're very, very keen on data quality.
It has to be good stuff. What we're starting with is survey data. We're not actually sitting here saying "people are watching this show."
We're constantly in the field. Tracking a whole bunch of different metrics, which were derived from the original analytical piece that we did. For example, we look at about, I think it's 28 different measures of satisfaction across each of 33 or 34 different subscription video services that we tracked.
With that data we're able to understand which satisfaction measures correlate more strongly with other outcomes. For example, likelihood to be considering churning, because we're also asking people not only "Which services have you signed up for or dropped in the last month?", but also "Which services are you considering signing up for in the next three months, six months, whatever it might be?" Then we're able to do a lot of analysis behind the scenes and understand there are certain satisfaction measures which correlate more strongly with intent to churn or intent to remain for that matter.
There are other measures as well, about satisfaction, which are also very similar. We ask a number of different questions about how people engage with a service. How many people within the household watch the service and so on?
How indispensable is it? We have a whole ranking of services on what we call indispensability, which if you look at it the other way is how disposable or not is a service?
Is it somewhere you go when you're first, when you're looking for something to view as second, third, and so on? All of those sorts of metrics we have, which combined give us a very strong perception of actually what matters and how it matters differently to different segments, different types of subscribers, those who are inclined to churn, those who are not inclined to churn, and so on. As it relates to different services, because to the point I mentioned earlier, not every service is equally central. I talk about service centrality a lot. It's central to somebody's entertainment lifestyle. It's not as important to them. We have to understand how all of those things come together. The key thing, though, is when Claritas, and Ron can talk about this with more detail than I can, but when Claritas ingests data from a source, they then model it.
We are looking at how many connection points do we have between our data and the data that they have to be able to model together to create a probabilistic model where you can actually say the segments of the population which exist within Claritas are the ones that best match the segments that we look at or the subset of an audience that we're looking at. We might, for example, get very, very specific and say we want to look for people to whom these three measures of satisfaction are disproportionately important in relation to the amount of use they get out of a service and their intent to churn because somebody wants to target that audience in order to reduce churn and perhaps bring people from another service where there's a level of dissatisfaction on those three metrics,. Then we can match that audience against the Claritas identity graph and say of the 68 segments that make up the PRISM segmentation of the population, this subset is the most important, the tightest match to that group, and they can be targeted. That's really, I think, what makes this work. Ron, would you probably add more?
PRIZM Premier is built to be lifestyle segmentation, and it's universal. It's everywhere. It's on the household lists. It's on the respondent data. It's on our own research.
It's on the Nielsen TV ratings. It's on Simmons. It's on MRI. It's everywhere, and we've created PRIZM-based audiences, digital audiences, and syndicated them, so they're out there on all the DSPs. You can buy PRIZM-based audiences without going through us, without going through Magid, because they're syndicated.
They're out there available everywhere.
Mike Bloxham: It's what links everything. I'll give you a really good example. That hypers group that I talked about, they're about 12% of the population.
They're that high-churn, high-value group. They tell more people about more content than any other group. They spend more money on entertainment-type expenditures, in a given month.
They're really important to you, even though they churn. Every streamer, in our view, should have a hyper strategy, whether they think of them as hypers or not. You should have a distinct strategy for these people that come and go, because your goal should be, if you're one of those services they're not going to stick with 12 months of the year, and that's true of most services, your goal should be to get another couple of months out of the year.
For one of those services (editors note: at StreamTV, he mentioned HBO Max), we did a calculation where we said, if they could get 25% of their high-churn subscribers to spend two months of the year more on the platform at retail, that would add $112 million to the bottom line. It gives you an idea of their importance, because this group, they're 12% of the population, but on average, across the last year, they've accounted for about 30% of new subscriptions, sign-ups every month, and about 21% of drops every month. So they are disproportionately important.
Claritas, with their system, is actually able to say which segments within their system, therefore which households, are most likely to be hypers. Then they can be specifically targeted, whether it's via email, social, open web, CTV, whatever it happens to be. And that's really important if you're going to put out differentiated messaging to this group, where you're trying to say, "We want you to come back a month earlier, we want you to stick around for a month longer."
And because we've also integrated another data set called emotional DNA, which is all about show titles, movie titles, content, that people watch. And an understanding of how they engage with it and how they perceive it in emotional terms. So we're able to overlay that so the messaging can be informed by saying, okay, people lean into this show, they watch it with their kids, they watch it because it's uplifting, it's moving, it's funny, it's outrageous, whatever it happens to be, we've got all of that.
So it can inform not only the targeting, but also the messaging to the specific clusters within Prism, the households, that are most likely to be those high churn subscribers. Just to give you that as an example of the sort of combined utility.
Ron Cohen: What's unique is that if you're talking about analyzing your subscriber file, a lot of them don't have name and address. The streamer doesn't know. They've got an email, or a device ID, or an IP address, or some other digital identifiers, and that's it.
That's all they got. We can take that, put it into our graph, because our graph is linking digital identifiers and postal identifiers, so we can apply these codes and build these profiles without knowing the PII of the people. We don't know their names and addresses, because they don't know their names and addresses.
They don't need them.
It's an interesting ability. From the TV side, there have been a lot of discussions that streamers need to "own" their customers. That having a direct financial relationship with them is key, because you can then get their data, their credit card number, their address, and it sounds like that necessarily isn't as big of a deal as people think it is. Not for this.
Ron Cohen: There's other reasons why you might want to own that data, right? That kind of data is pretty valuable to have, but you don't need it to do this kind of analysis, is our point.
Mike Bloxham: I think it's interesting. Yes, you can own data. You can have data, but really what one's talking about there is owning a relationship.
You don't own the consumer. I always think it's dangerous to talk in terms of owning the consumer. You don't own the consumer. You share a relationship with that consumer, and they have a great deal of power, and that's been made very, very clear since the advent of interactivity. People can control much more what they watch, when they watch it, and what they do, how they interact, and everything. Really, that's why streaming has emerged as a force in the landscape.
The important thing is that you understand enough about the subgroups within your subscriber base, and you're able to activate against that understanding to get commercial value out of it. That's what this is really about. Whether you actually need to know their names and addresses is another issue.
The reality is, businesses like Claritas prove you actually don't. You just need to know who you need to reach out to, where they are, and what you need to say.
Ron, I'm curious about what you see as the biggest challenge for you when it comes to collecting all these data points and assembling them. Because you're dealing with a lot of data, and if it's not aggregated and collected in just the right way, it's not going to be accurate.
Mike Bloxham: Or it's not going to be privacy compliant.
Ron Cohen: Yeah, that's the other side of it too.
There's a lot of that to it, and we've had over the last couple of years to strip data out of our identity graph in order to maintain compliance with all these new state privacy regulations related to sensitive personal information. A big part of our practice has to do with multicultural segments, since that's where most, if not actually all, of the growth is. Certainly the population growth and most of the spending growth is coming from multicultural segments.
We have a part of our practice that's devoted to identifying, segmenting, understanding, analyzing, communicating with multicultural segments. It's a huge part of the streaming universe. In some cases, we've been forced to now look at neighborhood level data, because it's no longer legal to possess that information at person or household level in certain states, and that list of states is growing every day.
Just this week, two more states were added to the list, and there'll be two more between now and the end of the year, and two more are already scheduled for January, so it's getting harder and harder. Between you and me, I don't think they're doing anybody any favors. Most of our customers are looking to see what language they want to communicate to certain customers in, particularly Hispanics, but not only Hispanics, and the states are making it harder for them to do that, which is not, I'm sure, the intended effect, but it is the practical effect.
Well, like most things, there's a lot of unintended consequences, too.
Ron Cohen: That's right. That's exactly what it is.
One of the many reasons I wanted to have this conversation is that there is so much of what your two companies to that people just don't understand.
In your case, Ron, a lot of people would worry, "Oh, they've got all of this data on me," without realizing it's not being used the way they visualize or the way they are afraid it's being used.
Ron Cohen: Well, of course, we're completely compliant with all the privacy regulations, and if you want to opt out, there's a website to go to, and you can opt out, forget me, don't use my data, don't sell my data, all of that. But these new regulations that I'm talking about are opt-in. In other words, you need affirmative, informed, opt-in consent to even possess sensitive personal information. Now, unfortunately, that's a real limited set of data. So yeah, it's getting harder and harder to do that, to collect it, to stay compliant, and to still be useful.
Mike, I'm curious when these discussions are happening, and maybe you can address this as well, Ron. Discussions are happening with customers about the data sets that you have and the insight that you can provide. What's something that surprises some of these companies you're talking to where they go, oh, I didn't realize somebody could do that?
Mike Bloxham: You mean our kind of streaming clients and so forth?
From the streamer side. Do they understand what you're capable of providing them?
Mike Bloxham: It's always very interesting when we sort of lift the hood on what we've learned and what we're doing. There's a couple of reasons behind it.
Obviously, given streaming services lean into their own first-party data, as they should, some perhaps rely on it a little too heavily because without other data, it doesn't necessarily tell you everything you might want it to.
My guess is Netflix would fall into that category.
Mike Bloxham: I don't want to name anyone. But for example, it's logical to assume that what people watch is a really good indicator of whether or not they're going to churn and what else they're going to watch and all this kind of stuff.
And to an extent, that is true. But actually, what people watch isn't a great predictor of churn. Those hypers that I told you about, those high churn people also watch a lot of the same content that some of the very stable subscribers watch.
So just looking at them from the point of view of what they watch isn't going to tell you whether somebody is high churn.
Now, if you've got household records over time and you can see that, yes, you've got that on an individual basis, but it's very difficult to generalize because you don't have the wealth of information behind that, beyond what they're viewing, to actually be able to model beyond data. Which is, again, where Claritas comes in as well as some of the other things that we have.
For example, we found, and this is one of the things that people are leaned into, that the strongest single predictor of churn is an attitudinal one.
And you don't see that in any behavioral data, in any first party data, because it's all up here (motioning to the head) and here (motioning to the heart).
And it actually comes back to that original question that we were asking about the likelihood of churn within a given period of time.
That's actually a very, very good predictor.
You would logically expect that it would be, but those sort of things don't always play out.
It doesn't mean that everybody who says that they're going to churn out in four months is going to churn out in four months at all, but it is one of the very good predictors.
There's a couple of other things that we have.
I mentioned this indispensability thing. We have a ranking of indispensability in a score for each of these segments by respondent. That is one element within a framework that we have, which is a strong predictor of churn or the intent to remain.
Likewise, this issue of what we call primacy. Do I go there first? Do I go there second?
Do I go there third when I'm typically looking for something to watch? How many shows do I watch on a service? Is it of value to more people in my household?
Do the kids use the service? There's a whole kind of matrix of different factors, some of which make intuitive sense, some of which are more surprising, which are important about that. That's one thing that people have leaned into because churn has become such an incredibly important thing.
Obviously, we are no longer in a marketplace which is just about organic growth with all boats rising. The number of net new subscribers coming in is now very, very, very much smaller than it was a few years ago. We actually have one segment within our segmentation who are out of market at present.
They're what we rather unkindly call the Inerts. Some of them dip in and dip out again. That group is very, very slowly diminishing in size, but they're not equally available to everyone.
If you're Netflix, if you're Amazon, if you're Hulu, maybe Disney because of the brand, then that's a group you can potentially spend money on successfully. If you're in the area, you get to the middle ground or the long tail, unless you've got a piece of content which is capturing an inordinate amount of interest in the population at large, that group is going to be much, much harder to obtain. This is a share driven market now.
It's not a growth driven market. One of the ways I characterize it, it used to be kind of a land of unicorns and rainbows where all boats were rising. Now, it's more like a knife fight in a dark alley.
It's actually much more like the telephony market where mobile providers are trying to steal a share from each other or the packaged goods market. Trying to get tiny incremental percentage points of growth while at the same time trying to get those people who are current subscribers to spend more time on the platform, to find more content that they want, to make them stick it, to make it less likely they're going to churn or if you're ad supported, of course, to create more inventory that you can monetize because more time on the platform means more inventory.
One of the other things that is advantageous about the data is we do look so deeply and so broadly. Like I said, we track 33, 34 different SVODs plus we do look a lot at the FAST networks as well and see how the relationship of use between people who subscribe to the likes of the Netflix,
Paramount+, all the rest, how they interact with FAST platforms. But the fact that we look so deeply across so many means that you can look at the data in many different ways and see where any given service sits relative to others and see where their true competitors might be.
For example, how subscribers of one platform are considering signing up for which other platforms in the next six months. That's really valuable information. Particularly then if you can look further and say, okay, subscribers of platform one are all looking at platform three.
More than, there might be like 18% of them are actually considering signing up for platform three in the next six months. Now let's look at how satisfaction with different aspects of our service as if we're platform three, match up against platform one. Now let's go target them.
Let's go target the subscribers of platform one with the messaging that emphasizes the quality of our offering relative to something they might perceive as being a bit weak on platform one. So the fact that you can play with the data in this sort of way is it gets very, very interesting for not only acquisition and retention, but for strategy. And because we have a lot of information tied in through our emotional DNA product, which is the product, you can actually get into programming and content acquisition strategies as well.
Ron: And ad sales, right?
Mike: And indeed ad sales.
Ron: We're collecting tons of information outside the streaming industry on where do you shop? What brands do you favor? What cars do you drive
How often are you buying a new one? What brand are you banking? Who are you insured with?
All of that is collected also and linked through Prism so that you can have a profile of what are the top four makes and models of vehicles that are purchased by viewers of this program. And to circle back and answer your question about what surprises most people in the streaming world, I think a lot of these platforms who don't have PII think that they don't have access to these kind of insights. And when we explain to them that we can take the digital identifiers, apply the segment codes, and open up not just our insights that are linked to those, but all the insights that Magid has to offer, that is truly eye-opening.
They're kind of staggered by the wealth of data and insights and analytics that then become available to them who have no PII about their subscribers.
Mike: And this applies to all the smart TV manufacturers as well, because they've obviously got their own services that they're running. And they have a similar issue. Who is using the glass and who is using their ad-supported streaming offerings as well.
And similarly, we can answer all those questions.
Ron: And all the FAST channels. And all the aggregators of the FAST channels. They're all selling ads, right? And we have great data to show them about what ads they should be selling against which program.
I have to say that generally speaking, for the television side of this equation, most people have no clue. Absolutely no idea what's available and what's possible.
Mike: Honestly, this is a lot of fun. We're learning things, we're doing things that have never been done before.