What is viral marketing?

From March to June in San Francisco alone there are three conferences titled "Viral Marketing," not to mention a few books being written on the topic. In some cases, the term is used quite loosely to refer to as simply "word of mouth" or "pyramid marketing." It also has been applied to a diverse set of entities--from "The Blair Witch" to Mahir.

So what is viral marketing? In 1997, when DFJ first coined the term in a Netscape newsletter, we used several examples to illustrate the phenomenon, without defining it beyond "network-enhanced word of mouth."

Its original inspiration came from the pattern of adoption of Hotmail, beginning with its launch in 1996. Tim Draper persuaded the company to include a promotional pitch for its Web-based email with a clickable URL in every outbound message sent by a Hotmail user. Therein lay one of the critical elements of viral marketing: Every customer becomes an involuntary salesperson simply by using the product. The product is fundamentally a communications product, and the marketing piggybacks on the message.

Viral marketing is more powerful than third-party advertising because it conveys an implied endorsement from a friend. Although clearly delineated as an advertisement, the spillover marketing benefits are powerful--much like the efficacy of radio commercials read by your favorite DJ.

The recipients of a Hotmail message learn that the product works and that their friend is a customer. A key element of consumer branding is usage affiliation: Do I want to be a member of the group--in this case, my friends--that uses the product?

We were amazed at how quickly Hotmail spread over the global network. The rapid adoption pattern was that of a network virus. People typically send emails to their associates and friends, whether they're geographically close or scattered around. We would notice the first user from an overseas university town, and then the number of subscribers from that region would rapidly proliferate. From an epidemiological perspective, it was as if Zeus sneezed over the planet.

Hotmail grew its audience from zero to 12 million members in 18 months, more rapidly than any company in any media in the history of the world. Fair enough--this is the Internet, after all. But it did so with an advertising budget of $50,000--enough for some college newspaper ads and a billboard.

Nonviral competitors like Juno Online Services spent $20 million on traditional marketing in the same time period with less effect. What's more, Hotmail became the largest email provider in several countries, including Sweden and India, where it had done no marketing whatsoever.

Hotmail is not an isolated incident. It and instant messaging service ICQ had close to the same number of subscribers at their six-, nine-, 12- and 18-month stages. What do they have in common? Hotmail was typically used as a secondary or personal account for communication to a close coterie of friends, much like ICQ's buddy lists. There appeared to be a mathematical elegance to their smooth exponential growth curves.

A first-order model for viral spread is this:

cumulative users = (1+fanout) ^cycles

In this model, the exponent cycles is the number of times the product is used in the time period since launch (or frequency * time). In the early days, Hotmail and ICQ fanned out to about two new members every month, and they each told two friends, and so on and so on. By the simple model, one seed user grew to three users at the end of the first cycle, nine by the second, 27 by the third, and so on.

Companies with much larger fan-outs, such as the free email list managers, have grown more quickly than Hotmail. Those that have provided an economic incentive to spam large groups, like AllAdvantage.com, which pays people to view advertising, have grown faster still, from zero to 750,000 users in two weeks.

The same formula would apply to traditional word-of-mouth marketing (like MCI Friends & Family discount plans and Tupperware parties), but lacking the involuntary coupling to patterns of communication, the average fan-out and frequency are much lower.

For a bit more accuracy, we can factor in the variables that describe the success of the recruiting message and the retention rate as percentages:

cumulative users = {(1+fanout * conversion rate) * retention rate} ^{frequency * time}

Working through the variables, the ideal viral product will be used to communicate with many people, will convert a high percentage of them to new users, and will retain a high percentage of them. It will also be used quite frequently.

A more accurate, second-order model would include decay functions on each of the variables, reflecting novelty and saturation effects. For example, Hotmail's variables are tapering as it reaches population saturation. Hotmail now has more than 75 million active users, which is one out of every four people on the Web worldwide.