While it is fun to throw about theories about what works and what doesn’t in viral marketing, the fact is that there is precious little public analysis about online behavior in the face of social economic incentives.
The most detailed analysis I’ve been able to find is “The Dynamics of Viral Marketing“, which looked at real data from an e-commerce site that provided a simple incentive–recommend what you’ve bought and you + the first person who buys off your recommendation get 10% off.
We present an analysis of a person-to-person recommendation network, consisting of 4 million people who made 16 million recommendations on half a
million products. We observe the propagation of recommendations and the cascade sizes, which we explain by a simple stochastic model. We analyze how user behavior varies within user communities defined by a recommendation network. Product purchases follow a ’long tail’ where a significant share of purchases belongs to rarely sold items. We establish how the recommendation network grows over time and how effective it is from the viewpoint of the sender and receiver of the recommendations. While on average recommendations are not very effective at inducing purchases and do not spread very far, we present a
model that successfully identifies communities, product and pricing categories for which viral marketing seems to be very effective.
The great thing about the analysis is that the data set extends to three major categories (books, CDs, DVDs and videos) with lots of buyers over two years. The researchers draw some interesting conclusions, including the observation that “purchases that resulted from recommendations are just a drop in the bucket of sales that occur through the website”.
Reading the analysis it becomes clear that the etailer did a poor job of setting up a successful social incentive system and of providing the tools to let consumers take advantage of it. For example:
- By not imposing limits on the number of recommendations that could be sent out they allowed some people to become spammers, which negatively affected conversion for the recipients, even as far as other senders were concerned. No surprise here–by analogy, email spam has reduced the effectiveness of email as a marketing medium for everyone.
- By having a one-size-fits-all incentive system, the etailer did not take advantage of the fact that the dynamics of viral distribution vary by product category and price range. Imagine picking a simple marketing strategy and putting it in place for two years without change and without any A/B testing. No, that’s not how things should be done.
- The etailer provided no tools to help viral spreading beyond the ability to email people with a recommendation. No profiles, recommendation lists, interest groups, checkout notifications, recommendation-based cross-selling, etc.
To make effective use millions of distribution channels, e-commerce sites need to think systematically about social commerce. They need a strategy first. Then they need to realize that implementing a successful social commerce system will require investment not unlike that required to build any major part of their site.
Great post… your absolutely right. There is quite a bit of “sayers” in viral marketing — but very little data to clearly define what works and systemizing it.
Wow! That graph of recommendation network propagation is awesome! Even if that were all this post mentioned about the study, I would read it that day.
btw. I agree with Maria.
well, i managed to finish it. jeez, there is a lot of material there; quite an in-depth analysis. one of the more interesting (and easier to implement changes because of) results is that “weak ties were important to the rate of information diffusion,” to oversimplify. Much of what I’d read in the past about viral marketing focused on the importance of those “supernode” individuals who were the minority which propagated product selections frequently. Those people are quite important, but there is a long tail of recommenders, particular for niche/subculture products. I’d always thought that, so it’s good to see some backup, finally 😉
Jim, right, the paper is full of little insights like that, e.g., what’s the most effective number of “forwards” one can make. Too little is not enough. Too many and people start thinking of you as a spammer.
Very, very interesting study – but nonetheless an isolated and non-representative case. As Sim commented: the propagation graphs are only as good as your campaign design. Poorly planned initiatives may result in conclusions like ‘Viral is all hype and no substance, it doesn’t work’. Wrong, of course. But ‘Viral’ is too many things to different people and certainly a lot broader than just primitive incentive schemes.
We need to keep in mind that ‘viral’ is not only (and not necessarily) about recommendations. In many cases it may be a ‘transport mechanism’, a many-to-many communication channel for a one-to-many message.
The effective, multi-connected nodes also don’t have to be seen as spammers – if they are encouraged to apply permission principles. There is nothing wrong in reaching many recipients, if they are happy to receive what they perceive as useful (or even just interesting) communication.
In a recent mobile telco case (that I am contractually bound to keep anonymous) we used a social network model to reach a large audience, by encouraging nodes to forward free (acceptable and desirable) content – in our case jokes. The only condition to subscribe to the free service was a permission to receive occasional product news from the source. After reaching a critical mass with just humour, the ‘product news’ was an irresistible special offer that was not visible in any other communication channel. (Unlike the company in the above study where ‘viral’ competes with a very visible site). As a result, the competition was puzzled by the sudden jump in customer churn and only figured it out when one of their directors (a keen subscriber to the joke service) received the special offer himself…
If we make hurried ‘wise’ conclusions like the one that viral sales are “just a drop in the bucket of sales that occur through the website”, then I can say the opposite: viral sales in our case were a Niagara compared to the drop that ocurred elsewhere.
While SNA is already a well established scientific discipline, its application in marketing (and commerce in general) is still in very early empirical (even cowboy) days. It requires cross-discipline competence and close collaboration between mathematics (propagation models etc) and the humanitarian disciplines that underpin marketing (and even higher business strategy), still a rather rare combination.
Which shouldn’t stop us all from trying 🙂
Pingback: Viral Dictionary « HighContrast