Paritosh Joshi: Ratings need reinventing

25 Jan,2013

By Paritosh Joshi

 

A story on this site published in May 2012, “TAM to cross 10,000 Peoplemeter mark soon”, signalled TAM’s intention to substantially deepen its coverage as India’s television footprint continued relentless growth.

 

It brought to mind a conversation I had with senior TAM personnel a few years ago where they explained to me the mammoth scale of the data processing task that tracking viewership involved. Here is a simplistic way of looking at it:

1 2 3 4
Homes Viewers (Age 4+) per home Average daily time spent (seconds) Unique data points (1. x 2. X 3.)
10,000 4 14,400 576,000,000

 

A single day’s dataset has very near 0.6 billion unique data points. Given that ratings are released weekly, the ratings tables that you read are compiled after compiling information from ~4 billion data points.

 

Let us now throw in a comparison with another medium we are all familiar with: Facebook. In September 2012, Mark Zuckerberg announced Facebook’s acquiring its 1 billionth subscriber. Over a half of these are active in a given week and post at a steady rate of 3 updates a day. That’s 1.5 billion updates a day or 10.5 billion a week.

 

In both cases we are talking about really large numbers. The difference is what happens next.

 

TAM crunches all the 4 billion data points down to 1 second granularity viewership trends for each channel that it tracks. That gives you, say, 400 channels being tracked. Facebook, taking a radically different view, starts trying to triangulate what are the likes, dislikes, interests and affiliations of each one of 1 billion individuals.

 

In the TAM view of the world, individuals are faceless, identity-less statistics who vote with their eyes for different channels and shows. In the Facebook view of the world, individuals are the very center of all analytical exercises helping the company offer individually tailored suggestions for everything from whom they should seek out to make friends with through what they ought to be buying.

 

The difference is telling. The legacy medium places the content at the centre of the analysis plan, the new age one, the consumer. While the first plan crunches a large dataset down to a relatively compact tabulation, the second embraces the concept of ‘Big Data’ where datasets going into the Exabyte order of magnitude (an Exabyte is 1 billion gigabytes) are routine.

 

Ratings have been around from times when mechanized data processing was in its infancy and the first task before any database manager was reducing and compressing voluminous data into a few large chunks that could then be subjected to analysis. In the specific case of television viewership, an easy was to construct a histogram that plotted the number of viewers against each channel and program. This histogram would then be projected up from the sample to the population to yield an estimate of the percentage of people who watched a particular program: the rating. Since this was the only way in which we had ever seen television viewership being tracked and reported we found nothing odd or inadequate about it. Even today, when digital media enable us to target individuals with very precisely defined characteristics, we still don’t challenge the rather coarse approach that ratings take.

 

So here is a thought: It is time for television measurement to place the viewer at the centre of the measurement system.

 

The advent of digitization in India’s television landscape throws up an interesting possibility. If a return path from subscriber to distribution platform is natively available, as it is in digital cable systems or is bolted on using various modes of internet access, as it is in DTH, it becomes possible to know continuously what channel the set top box is tuned to. Techniques like Data Fusion and Ascription (dealt with in a previous column that you can find here) make it possible to marry set top box data with respondent level Peoplemeter data thus magnifying it to large digitally connected populations, within defined levels of statistical error. It is now possible, provided we already have access to cable or DTH operators’ subscriber lists, to develop very good estimates of the viewership behaviours of individual consumers.

 

In effect, we can tell, within defined levels of error, what an individual in a digital cable or DTH home consumes on television through the day. We now have a view that is viewer centred rather than channel/programme centred. This is where the ‘Big Data’ approach must come in. The massive datasets that are born of the union of Peoplemeter and Set Top Box data need Big Data tools to be managed sensibly. Mining the datasets using these tools can yield an unprecedented level of textured understanding and individually addressable propositions.

 

And given that digital distribution platforms now have the ability to push messages and suggestions to the viewer, just like online media do, we can use such insights to deliver unique marketing messages, whether for broadcast content or for client brands.

 

Come on then, BARC, put that viewer at the centre.

 

Paritosh Joshi has been a marketer, a mediaperson and a key officebearer on industry bodies. He is developing an independent media advisory practice. His column, Media Matrix, appears on MxMIndia, usually on Thursdays

 

Post a Comment 

Comments are closed.

Videos