Putting people in the picture

Now, a single home shared by two people might have between them three phones, a laptop, iPad, desktop PC, big-screen TV with various peripheral devices – such as games consoles and digital video recorders – and also maybe a smaller more traditional TV. It is possible for all of these devices to be IP-connected in some sense, each spawning their own datasets, so there could be as many as eleven datasets relating to that one home.

But it is actually more complicated than that. A variety of content services might be available through each device, meaning several more datasets could be spawned and each of these datasets may exist in isolation – and be held by different organisations who either won’t share them (for business reasons) or can’t share them (because of data privacy concerns).

And yet although the datasets are isolated, the consumption of content and the use of these various devices substantially overlap with each other. So while it might be possible to hold a very detailed view of an individual service, without information about how everything else relates it’s only a very partial picture – and potentially misleading.

When data worlds collide
Over time the BARB TV audience measurement system has had to continually adapt in order to remain a comprehensive measure of viewing, and to cover the growth of new things like digital, larger screen TVs, PVRs, HD and on-demand viewing.

While BARB has also been pursuing new measurement via PCs and laptops, etc, the greatest long-term impact for television may be what can be done through the TV set itself, thanks to the growth of “connected television” or “smart TVs”. It now looks like virtually all new TV equipment (both sets and boxes) will have some form of capability for connectivity built in as standard, and it will be up to the consumer whether they choose to make use of it.

Connected TV will also have an effect on how people view commercials. They can be served, potentially, on an individualised basis, and so it is likely that viewers of the same on-demand programme will see different commercials.

So we’ve been searching for ways in which the panel-based measurement of connected TV can be improved. There are two main options:

1. The first would be for the meter on the TV set to gain direct access to data from the connected TV device. This would be similar to the type of measurement BARB is able to get from Sky set-top boxes, and we have approached platform operators and manufacturers about achieving this for other platforms over the past two years. Although this would potentially be of relevance to audience measurement around the world and not just in the UK, the lack of an immediate need from the manufacturers’ perspective seems to be a significant barrier to achieving this type of measurement. We’d like to see that change.
2. An alternative approach is to provide the facility for broadcasters or programme providers to add a measurement identifier to their video-on-demand (VOD) material. This could act as a label to identify whose content it is and (perhaps) which service it’s viewed through. The identifier will most likely be a watermark, in the form of a hidden audio code. The TV meter we’re using already has built-in capability to detect this, so we are exploring (with our supplier Kantar) whether this code is suitable for extensive use within VOD services.

It’s important to bear in mind that making progress on either of these approaches requires co-operation across the industry. Progression will facilitate clearer identification of the source of viewing – particularly for VOD or non-linear viewing – and so potentially open up a more direct understanding of the location of batches of server data within the bigger picture of all TV viewing. This then helps to resolve the issue of separate data worlds existing in isolation and yet (paradoxically) overlapping with data in other worlds.

Counting people
All the development priorities so far have been about measurement via a panel – based on a sample drawn from the population – because any measurement system that aims to produce information on what people are doing needs to be based on people in the first place.

But of course, a sample-based measure will only provide a usable measure for the content that attracts a substantial audience – not for every single item out there. For on-demand programming it may be more sensible to report aggregates across broadcasters, services, or programme genres. There’s also some caution that with the measurement techniques available, it may not be possible to provide exactly the same kind of reporting of commercial viewing that BARB currently generates for broadcast TV.

So while we are actively pursuing the development of the panel approach to audience measurement because of its great strengths, we do recognise the limitations. For audience measurement to remain comprehensive it is going to be necessary to make use of different types data – but sample-based measurement of people will need to be at the heart of this in order to properly understand, interpret and bring together new data coming from other sources.

This is an edited version of a presentation made to the 2011Television Symposium in Amsterdam on 3 November