By Paritosh Joshi
If you are reading this column with any professional interest, it is safe to assume you have done or been closely involved with one or more of the following things within the last year:
- Sold media inventory
- Bought media inventory
- Planned a media schedule
In any of these situations you would have to:
- Define the target audience
- Use widely used market research to assess and compare impact of the medium or media in consideration
- Price the medium or media as a buyer or seller or assess its or their value for money for the advertiser’s planned media expenditure
Inevitably, you would have to deal with television rating points, publication readerships, radio listenerships and the like. That’s where the fun begins.
With the target audience.
“Housewives SEC A and B, 5 lakh+ towns, UP,Bihar, Jharkhand”, one might say. “Men and Women, SEC A1, Top 6 metros” another might demand. Or even, “Women, SEC A1+, Mumbai and Delhi”. I have to add I am not inventing these, having heard them as specific asks or offers in situations I have been in close proximity to. To be sure, you could probably assign brands or media to all of them with not much effort. So far so good. It’s what happens next that makes no sense.
Someone with access to the right research will actually produce numbers purportedly accurate to within a decimal point to size said target audience and the extent to which a medium or combination of media will reach it.
This is bovine excrement, euphemistically speaking. Why, you ask?
Because all media research is based on statistical sampling, not a person-by-person census of every reader, viewer or listener of show or medium. Statistical numbers are estimates. They work on the twin ideas that all large populations are distributed according to the Standard Normal Distribution, the good old Bell Curve that we are all familiar with. Put simply, the notion that in any large enough group, there are a few thin people, a few fat people and a lot of people of intermediate weight (thereby making you wonder what happened to all of us in the Media and Entertainment fraternity, or whether there’s also an ABnormal Distribution to explain it). And that if you were to draw an adequately large random sample from this normally distributed population, the sample would retain all the statistical characteristics of the population such as Mean and Standard Distribution.
It can be shown that the minimum sample size required to ensure that the sample follows the behaviour of the parent population is 30. Samples of smaller size will exhibit asymmetries and other oddities of shape (things statisticians call measures of Skewness but never mind), that make them useless for drawing reasonable conclusions about their parent populations. As the sample available to extrapolate from becomes smaller, the error in extrapolation becomes larger, exponentially larger.
Thereby bringing us back to the issue of ratings and readerships and such. Take readership and the Indian Readership Survey for a moment. About 67 per cent of India’s population of 1.2 billion, ~160 million households are represented by just over 2.5 lakh respondents. Put another way, every respondent represents nearly 1000 households. Things get even more interesting when you look at television metering.India’s 130 million (your guess is as good as mine on what the actual number is) are represented by ~8,000 meters. Of course, TAM makes no claim to represent all India, so even if these 8,000 only represented the top 100 cities that have a 2011 population of 128 million or a population above the age of 4 of ~115 million people in over 20 million homes, there would still be only 1 meter in every 2,500 homes. We will get more generous and allow for the fact that TV penetration across the top 100 cities is 70 per cent. In other words out of 20 million total households, there are only 14 million TV homes. Even in this situation there is just 1 metered home in 2,000 TV owning homes.
You see where this is going?
As users slice and chop large aggregate populations and search for meaning in the samples that supposedly represent the segments thus generated, the available sample used to do the statistical prediction shrinks to a point where there is no predictive integrity within it. And yet, statistically naive people in every corner of our industry routinely use these frail foundations to build imposing edifices of brand and media transactions and planning.
Then again, even the Taj Mahal is built on flimsy marshland that may eventually cause it to sink out of sight.
So here’s the suggested Statutory Warning: “Irresponsible use of audience measurement may lead to impaired business diagnosis”.
Paritosh Joshi was until recently CEO, Star CJ. He has been a marketer, a mediaperson and a key officebearer on industry bodies. He can reached via the comments board below or his Twitter handle @paritoshZero.