Tilling new ground in mobile media measurement

We sat down for a telephone interview with the company’s vice president of marketing Evan Neufeld, a former ComScore man, to find out more about Ground Truth, its place in the US mobile eco-system and why 2010 is definitely the year when mobile makes its mark as a serious advertising medium.

So, how did Ground Truth come about?

Our founder Michael Libes was working at a company that was kind of a mobile search optimisation company and he was subscribing to data from companies like ComScore or Nielsen, and he wasn’t really seeing the kind of actionable, real-time clickstream data that he needed to make search work. What he saw were aggregate measures that were great for telling him things like ‘one of two people use SMS’ or ‘three in five people use non-voice services on their phone’ but any type of real granular information about browsing or application usage didn’t exist in those tools because they are all survey-based. He started to think about this problem, and then he had this ‘a-ha’ moment: what if we could go to a host of different partners, all these different people who collect data about what happens on the mobile internet, and aggregate that across from a market-level view, and report that data out at an appropriate level of depth and frequency, where people can make real tactical adjustments to their content and advertising strategies on mobile devices.

I’m guessing that the team at Ground Truth are firmly behind the view that 2010 is the year of mobile, despite similar claims being made for the past few years?

Actually, I’d be more specific than that. Three years ago it was the year of mobile video, than the year after that it was the year of mobile advertising. If you look at the percentage of people who are using phones for things other than voice communications, which is probably about 60 or 70% of the market right now, 2010 is clearly the year of the non-voice phone. Is it the year when lots of marketing dollars come online? Is it the year when smartphone penetration goes from 20% to 30% or 40%? It’s hard to tell – but it’s definitely the year when everyone understands mobile is a serious medium that really has advertiser attention.

So where does your company fit in? You’re looking to work with media planners and site optimisers in tailoring ads and content, is that right?

I’d say that’s a little further down the road. You need to just back up a bit. At the moment there are lots of companies that measure different things. There are analytics companies like Pinch that supply individual publishers with data; there are companies like ComScore and Nielsen that do survey data that shows aggregate trends. We’re another tool that I think is necessary for people here – that is, to provide a market-level view of a number of things: 1 ) how my site is doing against my competitor sites in terms of traffic, ad effectiveness, etc; 2 ) how my site is performing against other sites in my category, or other select sites; and 3 ) how can I optimise my site based on where my traffic comes from and where it goes – this has implications for media planning and selling, etc. Right now, its more a baseline thing: our last week of published data had over 1,000 sites and 70% of those have never had a third-party validation of their traffic because they simply don’t appear in survey-based tools like ComScore and Nielsen. For some of these sites, they can now go to media buyers and say “Hey, I actually have a highly-trafficked site where people spend a lot of time, you should spend some money with me”. That’s the baseline. Beyond that, because we collect data at the network level, we will also be able to get to that point where we are tracking the effectiveness and impact of ads that are served in these mobile browsing or application-based environments. We’re not actually doing that yet, so it’s kind of a two-pronged thing. Number one: lets get baseline measures so everyone understands the total width and breadth of the mobile internet, and number two: lets move that forward and help people actually optimise the advertising and marketing they are doing in those spaces.

So where does your data come from?

We can’t disclose exactly where we get the data from, but I will give you some hypotheticals. Anyone who actually gets data at an aggregate network level would be someone we could partner with, so that could be carriers, MVNOs [mobile virtual network operators]; it could be OEMs [original equipment manufacturers] and SMS providers; it could be application measurement companies, browser companies – anyone who is touching multiple consumers on multiple devices across networks. The only nuance I can give you there is that we currently partner with a diverse group of companies.

What’s the reason for the secrecy? Is it because the suppliers haven’t got permission from their users to give you the data?

No, that’s not it. We actually go through a pretty rigorous process and in most cases the existing terms and conditions of their user agreement allow them to pass on anonymised usage data. The thing I think that is a little difficult to understand is that we have a very different situation with the carriers in this country than anywhere else. Outside the US, through organisations like the GSMA you can get all the carriers in the room and they can agree to do certain things that are going to elevate the industry. You can’t really get that in this country. These people are hyper-competitive. Lets say we had data from carrier A and carrier B: it’s really important to be able to mask where that data came from to the marketplace because neither carrier would want anyone to have that sort of insight. You have to work very gingerly with any of these guys. Right now in this marketplace, we have to be totally anonymous about who is supplying us with data. Do we see a point in time when we would be able to report that? Possibly, but its challenging enough to get people to share their data. [NB: Our interview with Evan took place before we spoke to ComScore’s Paul Goode, who told us that discussions were underway with a view to replicating the GSMA Mobile Media Metrics model in the US.]

What sort of form does the data come in?

Basically, we get totally anonymised data that has all the personally identifiable information stripped out of it. That data is a record of a string of activities – similar to a log file – that are unique to a single individual device. That allows us to understand all the hits and calls to a server based on what a person does. Just like when you surf to a website, it logs how much time a person’s spent there, where they go after that, how long they spend on a particular part of a site. It’s analogous to Hitwise in terms of the type of data that we get, we just don’t get our data from ISPs.

Given all the disparate data sources you say you have, how do you make sure you are not duplicating traffic numbers? How can you be sure you’re not counting the same person three times across three data streams, for instance?

We have a process for de-duping. For carrier data it’s less of an issue – people don’t typically reside on multiple networks, right. In the case of other data providers, I can’t get in to too much detail but it really depends on the type of data you are getting.

Do you have similar problems as internet measurement does with cookie deletion/multiple cookies when it comes to accurately measuring web browsing?

It’s not an issue in mobile. We have other problems in reporting. I’ll give you another hypothetical: lets say I told you we had iPhone data. You’d assume that we either had information from Apple or AT&T, so every time a device is proprietary to a network we can’t report out on it, but that might change at some point in time.

What about the issue of demographics? This type of information is held by the carriers and I guess you can’t get at that just yet.

There are two questions here: Firstly, will we ever get that data from our data partners? I think not, and I don’t think we’d want that data. Second, is there a way from a research science standpoint to project the existing data and map that to demographics and market-level data? There absolutely is a way to do that. Hitwise has done this, where they take anonymised information from their data partners and fuse that with demographics and project that to a market level. I myself was involved in several projects in a previous life where you took one anonymous data set and fused that with a representative demographic sample. That’s definitely something that can be done. We currently don’t do that. We wanted to go to the market with clean, actionable data and see if the marketplace actually wants demographics on top of that. If they do, then we will go down that road.