Viral marketing and randomness

I give a lecture in my second year microeconomics course at King’s on irrationality. It’s not in the standard intermediate microeconomics sequence but I think it’s incredibly important. The lecture focuses on the role psychology plays in economics, which leads into the emerging field of behavioural economics.

Traditional, neoclassical economics assumes that “man” is rational – so called homo economicus. But for a variety of reasons we’re not. One reason is that we’re not very good at maths and, in particular, we’re not very good at estimating the likelihood or probability of an event, especially if it’s rare. For example, you’re more likely to die from a bee sting than in a plane crash. And you’re more likely to die from slipping in the bath than either of them.

According to Transport for London, the average London bus has 17 passengers, which will come as a surprise to most of us who use buses in London. But there are more people on busy buses to get the impression that they’re busy, and we’re more likely to remember to unpleasant busy commute than a leisurely, empty ride. Our experience is skewed or biased and consequently we make a mess of the maths.

There is a debate forming on Enterprise Britain between Keith and Kathryn about the (un)importance of social media for business. I don’t plan to weigh in on the pros and cons of digital marketing, but I want to talk about a related topic: viral advertising.

Like the Supreme Court’s definition of hard-core pornography, I don’t know what a viral is, but I know it when I see it. My all-time favourite is John West’s 2001 “grizzly bear” campaign for their Alaskan salmon. It was hilarious and I was quick to forward the link to everyone in my address book.

Understanding what makes a viral ad successful is important for marketing folk. A recent paper by Yahoo researchers attempts to unravel the characteristics of viral “tweets” on Twitter – so called “cascades” – and it appears to undermine marketers’ hopes of defining and predicting a successful viral campaign.

Firstly, they conclude they’re rather rare. Ninety per cent of tweets are never “retweeted” (forwarded/repeated), and the rest are generally only retweeted only by a person’s immediate followers, not by those at two or three removes from the originator.

Secondly, it appears almost impossible to predict which tweets will go “viral”. Knowing whether a user has previously started a cascade provides the most information on the chances of starting another, which statistically speaking is very little.

In that respect, viral adverts are like buses. Our experience is biased and we overestimate the likelihood of virals succeeding, falsely believing that they are predictable, because we’re only really aware of the successful ones. As Nassim Taleb would say we’re “fooled by randomness”.

The research needs to be extended to other social media, but the implications are clear: predicting successful viral campaigns is unlikely. At present, they look to be fairly random despite what ad agencies will have you believe.

Please leave a comment - we all like them