From Survey To Server

When you look at the shape of most media plans, they all look remarkably similar. This is largely because media choice is led by what data are most available. And by what data planners are most comfortable with.

But there is a fundamental change taking place with regard to 'media' data, in line with the changes taking place as a result of the new digital world. The sort of data we are most comfortable with as planners and strategists are gold standard media currencies. These are long established, committee led, peer-reviewed, stable (mostly), trusted sources. I'm talking BARB (TV), RAJAR (radio), NRS (print), and so on. You can also include TGI in this as it has almost become a currency in itself.

Whether they are accurate or not is another point. And to be honest it doesn't really matter that much. It's more important that they are stable, to allow trading to take place. When you lift the lid on some of these they are in fact creaking under the strain and almost quaint in the way they measure media. However, measuring media is a huge challenge and they are the best we've got (I wouldn't knock things like BARB, they are very professionally run).

However, the internet and mobile can offer a different type of data. Instead of measuring them via surveys, we are starting to see data emerge that is closer to a census, in other words just counting what happens. There has always been a discrepancy between web measurement surveys (ie Comscore, Nielsen) and server data. But these issues (around things like cookie deletion) are slowly being resolved so the two are starting to align. Comscore is now integrating the two with their web measurement and their mobile measurement is purely server based.

Of course the downside of moving from measuring the person to measuring the device is that you don't know much about the individual, or at least you have to make an educated guess.

The other problem with server data is that there is too much of it. It's bottom up, not top down. You get everything, when in fact you only want some simple metrics. And that is the stage where we are now. The data are there, but no-one has yet figured out a way to make sense of it. Interestingly, this also happened with cable TV in the UK. Everyone thought that measurement would be simple using set top box data. I remember a project with Videotron to try to make sense of their set top box data. It was killed-off as unworkable as there was too much data in too many different places. Looking back this should have come as no surprise - they couldn't even issue their customers with correct bills.

So we need to bring in some different skills to reduce this mass of data to something that a planner can use alongside the other media data they already know so well. Comparable metrics that allow a mobile app to be judged alongside a full page ad in The Times. Traditional research companies with skills in sampling and survey design might not have the ability to do this. On the other hand the tech guys don't understand what metrics are needed. Some hybrid of the two is going to be required. A sort of techy researcher - a genuine nerd.

How Much Is Enough?

The previous post about online audience measurement prompted a discussion in the office about the mechanics of measurement (heads up, bit of a nerdy post this one - the discussion moved swiftly to more pressing matters concerning Graham Onions and Steve Harmison).

Anyone who has worked closely with ratings data, be it TV, radio or print knows that the mechanisms for recording viewing/listening/reading can be complex. More importantly the definitions used are often no more than arbitrary. For those that are not familiar with these definitions these are the ones for BARB, Rajar and the NRS:

• TV viewing is defined as 'being in the room with the TV switched on'

• Radio listening is defined as 'listening for 5 minutes or more during a fifteen minute period'

• Newspaper reading is defined as 'reading or looking at for at least 2 minutes'

Beneath these definitions are more complex rules and regulations that also play a role in the process of measurement, eg the TSA areas for radio (that define where a station broadcasts), sort cards for print (that build the short list for the readership questions), and so on. But these definitions also have one other factor built in, namely persistence. You have to listen to a radio station for 5 minutes in order to be recorded as a listener. Two minutes is the rule for reading a newspaper. For TV it’s 30 seconds in a clock minute (consecutive). These are designed to weedle out those who aren’t really of commercial value to an advertiser.

So what is the persistence level for online? 5 seconds? 20 seconds? As far as I can tell there isn’t one (if anyone knows differently let me know). Imagine the scenario: you are sitting at home in front of your laptop and you click on a link. No sooner has the home page loaded than you notice your trousers are on fire. You shut down immediately and call 999. But you’re still counted as a unique user. Is that right?

The purpose of media measurement is to evaluate the likelihood that an individual is exposed to advertising. On this basis established media are getting a raw deal. There is no way the online world will give up audience by introducing a persistence level, so maybe established media should think about relaxing theirs.