Matt O’ Grady sits in a funny place. He tries to capture, in numbers, how much TV audiences watch in America. This could be on the regular idiot box, online, on their mobile, through their DVRs (digital video recorders) or on any other screen. It is a tough job and as cable stations and broadcast networks lose viewership to Hulu and YouTube, it gets tougher. Vanita Kohli-Khandekar caught up with O'Grady on the sidelines of TV.NXT, a television industry conference in Mumbai last week. Edited excerpts:
What are the big issues the US market faces?
Keeping up with wherever the content goes. So, following the content on the devices it could be on, monetising it - how to charge online, how much to charge if people watch TV on their mobile, etc. In fact, net metering is one-third that of the national people metering. So we have 25,000 households with national people metering, of which 8,500 have TV plus net monitoring.
What are the patterns in TV viewing?
In the US, time-shifted viewing is increasing and so is online. But TV is still the king, meaning a majority of the viewership still happens on the TV. Online TV viewing is still less than four hours a month (against more than six hours of daily viewing on the idiot box). Online viewing is not cannibalising TV viewing. Time-shifted viewing is a very serious trend — 40 per cent of DVR homes now use time-shifted TV, against 6 per cent a few years back. Since so much money is riding viewership — more than $70 billion — everything has to be precise.
How important is it to have reverse path data (from set-top-boxes/STB)?
Reverse path data could throw up surprises but it is not clean data. For instance, the STB could be on but the TV could be off. Only a hybrid approach can work in this case — say, an online panel.
If the sample is small, then it (reverse path data) helps complement the other data. So in the case of niche channels it would give a sense of who is watching what; it will not improve their ratings.
Sample size is a huge issue in the Indian market. Are there set ratios for sample to population size?
The key to good audience measurement is universe estimates. If a universe is 80 per cent of your demographics, then you need to add weights to correct the data for nuances.
How do you marry the viewing behaviour across media (online, DVR, mobile) to arrive at one metric which the advertiser can then use?
It is very expensive to do that and we are not a hedge fund (meaning, we will not invest in it unless the industry is willing to pay for it). We just announced an online (election) campaign rating along with Facebook.
