Demographics Meet Analytics: An Interview with comScore CMO Linda Abraham

Linda Abraham

Gimme a little background on Comscore

We began in 1999 and our goal at that point was to measure the internet. We recruited a panel of 2 million people – a million in the US and a million outside the US. They volunteered to join our panel in exchange for an incentive – any number of things. The most popular is we plant a tree in a 3rd world country for everybody who joins the panel. We’ve planted close to 2.8 million trees since we added thatincentive.

The people anonymously tell us their demographics and we verify that against other sources, which is an important part of the data quality process, and we can essentially, then, anonymously measure everything that they do. We measure search activity, video activity, their transactions in many countries, which websites they visit…so that’s kind of in our culture, in our history.

One of the things that has been very clear to us over this period of time is that people are very different from cookies. You have issues like cookie deletion, which are quite obvious and well known at this point. But there are a lot of other things that go along with counting people as opposed to cookies and digital interactions. In a household, Dad sits down to use the computer and then Mom sits down to use the computer and they both go to Yahoo…that’s one cookie, two people.

We have technology that helps us differentiate and understand that. By the same token, you have multiple devices per person, which may result in 7 or 8 cookies, all from one person. If you go to the same site with different browsers, that’s two cookies, one person. So there are a whole number of these dynamics, which are that important when measuring an audience. We have learned a lot about that over the last two years in particular.

We now get census measurements from the tag and we combine that with everything we know about panel data – the dynamics, the engagement, the ??? “best breed” kind of measurement. And the reason that is helpful is that panels have inherent weaknesses, for example it’s difficult to get ???? IBM, for example, is not going to allow their employees to join a panel so therefore you’re missing that measurement and in the event that your site garners a lot of at-work usage, that’s difficult to measure. Niche-oriented sites might not be represented in our panel, which is designed to be representative of the total US population. Niche-oriented sites are longtail sites, which statistically you get a lot of noise around your ability to estimate those, and so this approach, which we call Unified Digital Measurement, basically accounts for that total usage and kinda fills that gap and allows us to take the richness of the panel data and ???? the consumer, which addresses all these other issues – all the cookie issues.

So that’s what brought us to the point where we saw that when you put these together you have a whole other view of this landscape, and made it make a lot of sense for us to enter this space because we’re not interested in just bringing another analytics tool to market, we think that some of the fundamental technology behind this particular product as well as combining with the audience level measurement is really what differentiates it, because the world is moving so quickly now, so when you have access to understanding your audience, you’re changing your site, you’re changing your strategy very quickly, you need to know, for example, if you make a change in your transaction process, have you all of a sudden made it more difficult for older people who don’t like changes much, and to the extent that if they’re in your target you’re ???

So, this allows you to change dimensions on the fly. You can decide that it’s important for you to be able to understand ???? and those that didn’t because you’ve got different content patterns that you want to capitalize on and this allow you to create those dimensions quickly on the fly so that as you’re ??? and as you’re tweaking your doing course correction and you don’t have to back and say, “Let me get that new dimension ??? it’s all very very realtime.

>>How did you find a technology to provide these new capabilities?

We looked very carefully at this space and what we liked about Net Stat was the fundamental access to the disaggregate data as part of this atomic platform, as we’re calling it. It doesn’t make any assumptions about dimensionality going in ???? in other products, and because we know a lot about big data sets. With two million people you can imagine…it’s one of the biggest databases in the world, and so we have a lot of engineering expertise that – with that fundamental capability – allows you to do pretty interesting things with the data, pretty quickly, and that really taps into a core competency that we have at Comscore, and we see the power in that because as we look forward, and there’s a proliferation of devices that people can access, if you really want to understand cross-platform usage, at the person level, and understand that you’re developing a relation with the whole person ???? that the dimensionality of that has got to be fluid – 3 screens, it may be 4 screens, 5 screens, who knows – and also the type of content – we need to integrate mobile and video streaming and web page content – into this same reporting system has got to be really critical.

So, what we like about the NetStat technology is that there were no a priori assumptions about any of those dimensions going in. It is a technology that lends itself to where we see the industry going in terms of flexibility of analysis.

>>I imagine that over the last 11 years that the demographic mix on the Internet has changed a lot and continues to.

It has, it continues and what’s really surprising about how it’s changing is that some of the behaviors that you wouldn’t expect are really taking hold in certain demographic groups. One of the examples we looked at recently is that we did a study looking at how women use the Internet differently from men, and looked at it on a worldwide basis, and if you look at ages of women – younger women, older women – of course you find that on an absolute basis there’s a much higher number of younger women doing social networking and fewer older womeen. But, if you look at the growth rate, more older women are getting into social networking. And when you look at engagement – how much time they spend – it’s actually just as high as the younger women.

>>I’ve noticed that my cohort – the boomers – were slow to jump in, but just in the past few years friends I’ve been trying to bring online for 20 years are all appearing on Facebook.

Yes and the demographic shifts are really important to understand because they represent new opportunities. Two years ago you couldn’t reach older women on the Net. Gaming is another example – one of the fastest growing segments of older women can be found in gaming. Not the shoot ’em up kind of gaming, but what they call “casual gaming.” That is rapidly taking hold among women. So there are lots of niches like that surprise you in terms of how people are using the Internet and the opportunities that that speaks to that you wouldn’t necessarily know unless you really studied it and

At last week’s eMetrics Marketing Optimization Summit in San Francisco, I interviewed Linda Abraham, comScore‘s Chief Marketing Officer about their recently announced acquisition of the Dutch analytics company Nedstat.

SDC: Can you give me a little background on comScore?

Abraham: We began in 1999 and our goal at that point was to measure the internet. We recruited a panel of 2 million people – a million in the US and a million outside the US. They volunteered to join our panel in exchange for an incentive – any number of things. The most popular is where we plant a tree in a 3rd world country for everybody who joins the panel. We’ve planted close to 2.8 million trees since we added that incentive.

The people anonymously volunteer to tell us their demographics and we verify that against other sources, which is an important part of the data quality process, and we can essentially, then, anonymously measure many things that they do. We measure search activity, video activity, their transactions in many countries, which websites they visit…so that’s kind of in our culture, in our history.

SDC: What kinds of limitations have you encountered using your panels?

Abraham: One of the things that has been very clear to us over this period of time is that people are very different from cookies. You have issues like cookie deletion, which are quite obvious and well known at this point. But there are a lot of other things that go along with counting people as opposed to counting cookies and digital interactions. For example, in a household, Dad sits down to use the computer and then Mom sits down to use the computer and they both go to Yahoo…that’s one cookie, two people.

We have technology that helps us differentiate and understand that. By the same token, you have multiple devices per person, and you can count 6 or 7 or 8 cookies, all from one person. If you go to the same site with different browsers, that’s two cookies, one person. So there are a whole number of these dynamics, that are really important when measuring an audience. We have learned a lot about that over the last two years in particular.

We introduced a new form of measurement just using the panel. We now have sites participating tagging with us, so we basically get census measurements from the tags, so we combine that with everything we know about panel data – panel dynamics, engagement, demographics, etc. and have formed kind of a “best of breed” form of measurement.

And the reason that is helpful is that panels have some inherent weaknesses. For example it’s difficult to get measurement of at-work usage from very large companies. IBM is not going to allow their employees to join a panel, so therefore you’re missing that measurement, and to the extent that your site garners a lot of at-work usage, that’s difficult to measure. Also, niche-oriented sites might not be represented in our panel, which is designed to be representative of the total US online population. Niche-oriented sites are longtail sites, and statistically you get a lot of noise around your ability to estimate those. So this approach, which we call Unified Digital Measurement, basically accounts for that total usage and fills that gap, allowing us to take the richness of the panel data in terms of understanding the consumer, which addresses all these other issues – all the cookie issues.

So that’s what brought us to the point where we said, well, you put these together and you have a whole different view of this landscape, and it made a lot of sense for us to enter this space because we’re not interested in bringing just another web analytics tool to market. There are a lot of them out there and people are happy with them in a lot of ways.

We think that some of the fundamental technology behind this particular product as well as combining it with the audience level measurement is really what differentiates it, because the world is moving so quickly now, that when you have access to understanding your audience – you’re changing your site, you’re changing your strategy very quickly – you need to know, for example, if you make a change in your transaction process, have you all of a sudden made it more difficult for, maybe, older people who don’t like changes much? And to the extent that they’re in your target, that’s a problem. You need to know about that.

So, this allows you to create dimensions on the fly. You can decide all of a sudden that it’s important for you to be able to understand conversion activity between the people that authenticated vs those that didn’t because you’ve got different content patterns that you want to capitalize on and this allows you to create those dimensions very quickly on the fly so that as you’re learning and as you’re tweaking, you’re doing course correction and you don’t have to go back and say, “Let me talk to somebody and get that new dimension implemented.” It’s all very very realtime.

SDC: How did you find a technology to provide these new capabilities?

Abraham: We looked very carefully at this space and what we liked most about Nedstat was the fundamental access to the disaggregate level data as being part of this Atomic Platform, as we’re calling it. It doesn’t make any assumptions about dimensionality going in, there are no limits on the dimensionality as there ar in other products, and because we know a lot about big data sets.

With two million people you can imagine…some of the biggest databases in the world, and so we have a lot of engineering expertise that – with that fundamental capability – allows you to do pretty interesting things with the data, pretty quickly. That really taps into a core competency that we have at comScore, and we see the power in that because, as we look forward there’s a proliferation of devices that people can access. And if you really want to understand cross-platform usage, at the person level, and understand that you’re developing a relation with the whole person and their day-to-day life, the dimensionality of that has got to be fluid. Today it’s three screens, it may be 4 screens, 5 screens, who knows – and also the types of content – so our need to integrate mobile and video and streaming and web page content into this same reporting system has got to be really critical.

What we liked about the Nedstat technology is that there were no a priori assumptions about any of those dimensions going in. We concluded that it is a technology that lends itself to where we see the industry going in terms of flexibility of analysis.

SDC: Certainly over the last 11 years the demographic mix on the Internet has changed a lot and continues to.

Abraham: It has, it continues and what’s really surprising about how it’s changing is that some of the behaviors that you wouldn’t expect are really taking hold in certain demographic groups. One of the examples we looked at recently is that we did a study looking at how women use the Internet differently from men, and looked at it on a worldwide basis, and if you look at ages of women – younger women, older women – of course you find that on an absolute basis there’s a much higher percentage of younger women doing social networking and fewer older women. But, if you look at the growth rate, more older women are getting into social networking. And when you look at engagement – how much time they spend – it’s actually just as high as the younger women.

SDC: I’ve noticed that my cohort – the boomers – were slow to jump in, but just in the past few years friends I’ve been trying to bring online for 20 years are all appearing on Facebook.

Abraham: Yes and the demographic shifts are really important to understand because they represent new opportunities. Two years ago you couldn’t reach older women in social networking. Gaming is another example – one of the fastest growing segments in gaming is older women. Not doing the shoot ’em up type games, but what they call “casual gaming.” That is rapidly taking hold among women. So there are lots of examples like that surprise you in terms of how people are using the Internet and the opportunities that that speaks to that you wouldn’t necessarily know unless you really studied it.

SDC: Well thank you, Linda, for helping us understand more about the dimensionality of analyzing personal behavior on the Web. And good luck at the conference.