FEATURE12 March 2010

Getting the measure of viral ads


We speak to Duncan Southgate, global innovation director at Millward Brown, about the firm’s efforts to gauge whether ads will go viral.


Research: How did you come up with this measurement?

Duncan Southgate: Millward Brown has an enormous database of TV ad pre-tests with well over 50 measures of the ways people talk about ads, so we’ve got a really good understanding of how people react. In the UK about two thirds of all the ads we’re testing are going up on YouTube, and it’s about one in three in the States. It seems fairly common practice now that most ads are given the chance to go viral. The slightly unfortunate thing is that very few actually do.

So we tracked down 102 ads on YouTube, captured the number of viewings online to get a views per week measure, and correlated that with the creative attributes from our Link pre-test. The next step was fairly easy: once we’d identified the shortlist of variables that had a relationship, we worked to pull together the best algorithms.

So what’s the key to viral success?

We came up with a slightly cheesy ‘ABCD’ framework: awareness, buzz, celebrity and distinctiveness. What we’re saying is, when you pre-test your ad to see if it’s got viral potential these are the responses you need to look for. We see viral ads that are really funny or exciting or gripping. Sex sells on the internet and we’ve certainly seen a few examples of that. The other emotional response that we’ve found correlates very strongly with viral viewings is surprise. Surprise and excitement work very well, and then there are specific viral tactics like ‘Is it real or fake?’ which always seems to generate a decent amount of interest virally.

What do you advise clients to do once they have their score for viral potential?

The advice, more often than not, would be that this isn’t worth spending too much because you could spend a lot of time trying to make it go viral and it probably won’t. That will be the story that most clients will actually get. But the ones that are lucky enough to have an ad that is genuinely exceptional and has got the potential to go viral, the question to ask is, who is it most likely to go viral amongst, so that when they do their viral seeding they can get some clues as to the right place to start the fire. Seeding widely is generally going to be the right advice, because the more places you drop this the more chance it’s got.

When things are looking good you should absolutely be making the most of it. You’ve got the potential to waste your time and effort if you’re the wrong side of the line, but once you’ve managed to get over the bar, seed it and flog it for all it’s worth.

There are other factors beyond viral seeding and online viral promotion that we’re sure have a significant impact as well. I think what’s changing in the industry is that people are planning for a viral element of their campaign in advance in terms of the creative they’re generating, in terms of the way they’re planning their media strategy and in terms of the type and style of creative they’re coming up with in the first place. The campaigns that do really well virally, like the T-Mobile dance, they’ve thought through the viral potential of that right from the outset.

Isn’t it risky making predictions about something so fickle?

I think it’s all in how it’s presented, because as long as people understand the limitations of our research, it’s better to have that research understanding than not. Your current option is to guess how other people will respond based on your own response. A pre-test can give you a representative response from 150 consumers – that’s the value of it. We don’t predict a precise number of views, we tell clients how their ad does in relation to other ads tested on these measures, and we give a very broad range of prediction to make it very clear that the number of views they could achieve has a lot of variability.

I don’t think we’ll ever predict it as well as we predict in-market TV impact but I’m sure we can improve the validation we have. TV is a much more defined, predictable medium – you know how much you’re spending and how many impressions you’re getting. Virally there are so many other factors that can play a role.

Do you worry at all that too much research and testing could take the charm out of viral video?

Not really, because ultimately consumers have more control in this environment than any other. They’re choosing to share these ads, all we’re doing here is helping clients identify whether their ad is sufficiently good that that consumer effect is likely to happen. Viral video exposure certainly interrupts consumers a lot less than some other ad formats out there. It’s about adding understanding to an organic process, not trying to fundamentally undermine that fun, interesting, live, organic environment.

1 Comment

14 years ago

Duncan - the ABCD framework IS a bit cheesy because Buzz is one of those self-fulfilling things. High Buzz = Probably go Viral. That's what buzz is. One thing you might look for is whether the TVC has a fresh, entertaining, signature moment - what I call the Hair Gel moment, in honour of Something About Mary. That moment was the thing that everyone talked about when the described the movie - it was funny, it could be explained in a narrative fashion, and it was genuinely fresh. Now in hindsight it may score high on your ABCD, but those high scores per se, don't explain why the scene was such a talked-about thing: they rather reflect the fact rather than predict it. I think you guys are barking up the right tree - but you're making it hard for yourselves by sticking to dry quantitative measures. One thing about things that go viral, they often go viral precisely because they don't follow a formula. Definitely watch this space though. Best regards.

Like Report