imrb international

In the time that it took IMRB to launch its Web Audience Measurement (WAM) system for analyzing usage of the Internet in India – we’d reported about its plans two years ago – much has changed: Comscore appears to have gained acceptance in India, Komli Media launched Vizisense and later introduced a freemium model. Nielsen hasn’t yet launched its own audience measurement product, and the space for an authoritative audience measurement system in India is still up for grabs.

WAN does a few firsts – it introduces the concept of WRP’s, or Web Rating Points, tracks time on site and “exposure”, which will be of relevance to display advertisers, and is also able to track instant messenger usage. But, it has its limitations. We spoke at length with Balendu Shrivastava, Group Business Director, IMRB, who has spent over two years on this product.

MediaNama: How does WAM Work?
Shrivastava: WAM is a true measurement of the Internet audience, and we’re using Panel information. Through the panel, we’re completely capturing the entire Internet usage that a person does, across browsers. It is not limited to Internet Explorer, Firefox or Chrome alone, and tracks Internet usage across browsers and across messengers, because we’re not embeding a software on the browser. It’s an exe file given to a panelist, and he just has to run that file. Every time he starts his computer, we start getting the data from the panelist. Our software is a completely independent entity, present in the network layer, pulling in data from there. Anything going to the Internet, we will tap into it.

MediaNama: With the panel, aren’t you missing out on the office usage? Will a user install this in his office computer? What about cybercafe usage?
Shrivastava: The moment we think about offices, we start to think about Infosys and TCS. If you look at it in terms of number of establishments with computers and Internet, 80% of these are SME’s. So I completely accept that my software will be thrown out by large companies, even though it is completely secure, verified by Verisign, and follows the WPP privacy policies. I still understand that this kind of software would not be placed in a banking environment or a very large software company.

MediaNama: But what about multiplicity of usage being lost, if you have someone on your panel who uses on a desktop at home in the evening, and an office PC during the day?
Shrivastava:We encourage installations in both cases, and we’ve seen instances of multiple IP addresses and multiple machine IDs where they are logging from. So people are accessing from multiple points. We also see weekend usage dropping, and picks up during the weekdays and worktimes. So there’s no reason to deny that we’ve got office machines. We’ve seen some installations ourselves at SME’s.

MediaNama: How have you done the panel selection?
Shrivastava: We’ve followed a traditional research methodology for that. We’ve broken down the sampling, as per age, gender and SEC classification. We have divided into 12 odd cells, and each city has 12 cells that have to be filled up. The recruitment is done offline and online. Even in case of online, people who have been recruited in the panel, each individual has got a call and has been verified. There’s an offline verification system.

MediaNama: How do you account for the self selection bias? (ED:users who apply online are likely to be mature Internet users)
Shrivastava: We’re not allowing any cell to have a very large number of people. Also, the sample is not only an online selected sample. Thirdly, it’s a completely verified sample, and we’re looking at other data – frequency of access, what time is spent, and we’re ensuring that we get a fair amount of spread, although that is not our sampling criteria, but a verification criteria. Self sampling bias comes from people who are expected to use the Internet a lot more. We’re monitoring these data points as well, in terms of what the response is stating, and we’re able to control that behavior. Not all people who submit data to me are taken into account. Over 50,000 people have registered and downloaded my software. I’m rejecting most of them, and I’m tracking around 6000, who are giving data to me regularly, according to demographic profiles that I need.

MediaNama: But 6000 is a fairly tiny sample. Vizisense claims to have over 60,000.
Shrivastava: They can do any number that they want to. We don’t know what kind of weighting they’re doing. We have a universe in place. TAM (Television Audience Measurement), for example, has around the same sample, for almost three times the audience. Secondly…

MediaNama: Questions have been raised about TAM’s accuracy too…
Shrivastava: TAM is being done by us as well, so we know what it is about, hold it to our hearts and say that that is something very accurate. What we have is an elaborate establishment survey (IAMAI-IMRB I-Cube survey), covering 90,000 individuals. I don’t think Vizisense has anything of that sort. That places credible estimates. We’re the only agency in India that can give you data as per that particular cell that we’re looking at. If I’m saying 15-25 male in Delhi, it’s the same establishment survey that I’m talking about. A lot of my weighting characteristics are taken of. I don’t need a self-weighting sample, which is typically a very very large sample size. I have achieved stability with a much smaller sample of around 2000. If I look at it from a sampling error point of view, a 6000 sample is almost reporting at a 99% accuracy level.

MediaNama: Again, there’s no way of judging whether it’s 99% accurate or not, because you don’t have a census to compare with.
Shrivastava: No, I have a universe figure, coming from the I-Cube (IMRB-IAMAI I-Cube Survey)

MediaNama: What about representation of sites? Unlike TV, the Internet is a larger universe. Wouldn’t there be an issue of availability or accuracy of metrics for a smaller sites, given that your sample is limited?
Shrivastava: I understand that there is an entire ecosystem out there. From a measurement perspective, we have to look at it from a cost-benefit point of view. The ecosystem of small sites, is currently mostly working on performance basis. It’s your top 20-100 sites that are doing advertising on a display basis, fueling brand initiatives. The smaller sites come in for lead generation. I understand there’s a need for covering small sites, but I’ll have to spend quite a bit to get there. The idea is to get a system in place which at least gets 80% of the online display advertising business in place, and then start looking at the remaining 20%.

MediaNama: Are you planning to expand your base of panelists?
Shrivastava: I’m currently reporting the top 8 cities individually, and the rest are covered as Rest of India. We’ll be expanding it, and taking smaller cities into the mix.

Vizisense Launches Paid Features In India; Appoints Australia MD
Nielsen Online Arrives In India; Clarity Or Just Adding To The Confusion?
Komli Launches Vizisense Audience Measurement System; Targets A Panel Of 50-100k
IMRB Plans Internet Audience Measurement System For India