Combining K-Factor Viral Marketing with Local Cascade Models
Sources: https://www.forentrepreneurs.com/lessons-learnt-viral-marketing/ and https://www.forentrepreneurs.com/the-science-behind-viral-marketing/
The basics of viral marketing is a simple exponential function (discounting retention rates): the number of customers at any given time is a function of the initial user pool multiplied by a K (viral) factor raised to how many time cycles has passed. What this implies is that for a product to achieve viral status, it must have a K-factor greater than 1. To simplify, this means that within each time cycle, an existing user will spread the product to K new users, and thus, if within each cycle, a user spreads the product to more than 1 new user, the product will achieve viral status. The K-factor is usually determined as the product of the number of invites a user will send and the acceptance rate of any invitation.
In a cascade model, we analysis the propagation of a product on a local basis. This works by examining a socially driven product, and each node in the network makes its decision to adopt the product based on how many neighbors adopt it. This establishes a framework for network cascades, in which when a certain fraction (q-factor) of a node’s neighbors adopts the product, the node will begin using it as well, which goes on to trigger other neighbors to also use the product. To model a more realistic cascade, however, several new factors must be introduced: heterogeneous thresholds and tie strength. Heterogeneous threshold is the differing of q-factors between different people; some people may be very willing to try something new, whereas others are not. Tie strength also factors in for this cascade model because an invitation through a strong tie is more likely to convert than through a weak tie. Finally, because of these factor, the local networks will form clusters, in which is a group of nodes wherein if even one of nodes begins to use the product, the rest of the group will follow. Clusters are usually comprised of dense, strong relationships that have low inter-relational thresholds.
To combine the K-factor model and the cascade model, we can think of each cluster as a node, for the reason that converting one member of a cluster is the same as converting the entire cluster. Between clusters exist most likely weak ties, in which there is only a chance of converting if an invitation were sent from one node to the other. From this model then, the K-factor would then be the number of clusters that a cluster can convert per time cycle. Furthermore, the K-factor would also need to be a function of the q-factor in addition to the existing # of invites and acceptance rate factors, because each cluster will now consider whether the neighboring clusters are also using the product. The number of invites now should also be adjusted to inter-cluster invites, as invites between members of a cluster does not contribute anything new to the viral status.
To summarize, in combination of the two models, there are some new factors to consider in viral marketing. From the existing K-factor model, important factors include the viral cycle time (determining how fast each iteration is), the number of invitations sent (namely the ones sent to the outside of each cluster), and the acceptance rate of each invitation. The invitation factors are mostly network-independent because it is essentially a cold invitation: the receiving party only has a weak tie and will judge the product based on its functionality. Added from the network cascade model, the q-factor will determine the size of each cluster; the lower the q-factor, the larger clusters will be, which will ultimately cause larger cascades.
We can see now that to create a viral product, the K-factor must be maximized and q-factor must be minimized. K-factor maximization involves features that draw in new users such as high intuitiveness and immediate usefulness and q-factor minimization depends on increasing the social benefits that stem from neighbors using the product.
Combining these two factors (and cycle time), we can see how existing social platforms such as Facebook and Twitter became the giants they are today.