An alternative to TAM for TV viewership data?

I've never been able to understand how media planners take TAM viewership data seriously. As per the 2011 Indian Census, about 50% households in India have a TV set.  Even if we take the 2001 data [1] for total number of households we would end up with 125 million TV sets in India. (the 2011 data is still being compiled, but its safe to assume that this number would have only grown).

TAM viewership data is based on a sample size of 3,454 out of those [2]

A sample size of 3,454 out of 125,000,000 TV sets.

I want you to back-up and read that line again. Digest those numbers.

Leave alone statistical significance, in what world would a 0.0028% sample give even directionally correct data? Thousands of crores of advertising money every year is being budgeted & spent based on data collected from such a small sample size.

If that wasn't enough, it turns out that they've been accused of corrupting & manipulating this data in favour of TV channels that are willing to cough-up some extra money.

Compared to the online advertising where there is probably too much data, offline advertising is like shooting in the dark. During the planning stage, you don't know who is seeing what channel to any degree of accuracy. After the ad-campaign you don't have any easy way of knowing how many people saw your ad (reach) & how many times (frequency). In fact, you have to pay more money to collect an inaccurate estimate of reach+frequency data.

There has got to be a better way.

I'm surprised why DTH/set-top box operators haven't jumped at this market. Today, you have an advanced mini-computer (in the form of a set-top box) attached to the TV sets of a very large number of homes.  Some of those set-top boxes already have the capability of recording 60-80 hours of TV footage. Wouldn't it be trivial to record the exact viewing behaviour of a particular TV & tie it to subscriber profile data (residential locality, income group, mobile number, channels subscribed, etc) 

As a media planner, pre-campaign, you would have access to up-to-date reports on which channel was watched, by what kind of household, at what time, and for how long. Post-campaign, you would be able to combine this report with the exact time slots when your ad was aired to easily compute reach & frequency. 

DTH operators could offer a discount to people who opt-in for letting them collect this data per-week or upload it on-demand via a SIM-card embedded right into the set-top box.

Does such a thing already exist? What would it take for Tata Sky & Airtel Digital TV to come-in and capture this market?

Update: It seems that people have already started moving in the direction of using STBs (set-top boxes) to record viewership data. However, the shift has been very slow. Sample this proposal by TAM to digitally quantify viewership and a company called aMap who seems to be getting into partnerships with Tata Sky and Dish TV for data collection.

PS: I'm not a fan of panel based data collection. Even in the online world, I am skeptical of systems that compile data based on opt-in panels. Here's my post about different traffic comparison tools.

[1] http://censusindia.gov.in/Tables_Published/H-Series/H-Series_link/S00-001.htm

[2] http://www.tamindia.com/tamindia/Faqs/India%20Peoplemeter%20Update%20-%20II.doc Section I(c) on page #2