This is an extension to my original three post series on user acquisition.
Over the past few months I been fortunate enough to give over a dozen talks at various events and companies about user acquisition, virality and mobile distribution. One of the best parts of the experience is that, without fail, every talk yields a new set of questions and insights that help me learn and refine my own thinking on distribution & growth.
One of the most common questions I get is around the difference between my definition of “viral factor” and the semi-standard definition of “K Factor” that has been floating around for a few years.
What’s a K Factor?
Wikipedia offers a fairly concise definition of a K factor, a term borrowed from epidemiology.
i = number of invites sent by each customer
c = percent conversion of each invite
k = i * c
As the wikipedia article explains:
This usage is borrowed from the medical field of epidemiology in which a virus having a k-factor of 1 is in a “steady” state of neither growth nor decline, while a k-factor greater than 1 indicates exponential growth and a k-factor less than 1 indicates exponential decline. The k-factor in this context is itself a product of the rates of distribution and infection for an app (or virus). “Distribution” measures how many people, on average, a host will make contact with while still infectious and “infection” measures how likely a person is, on average, to also become infected after contact with a viral host.
What’s a Z Factor?
Based on this framework, the Z factor is literally the percentage of users who accept a viral invitation that they receive.
The Problem with K & Z Factors
I meet with a startup that told me proudly that they had measured the viral factor of their new service, and that it was over 2. My first question, of course, was:
“over what time period?”
In my blog post on viral factor basics, I define a viral factor as follows:
“Given that I get a new customer today, how many new customers will they bring in over the next N days?”
The key to understanding viral math is to remember a basic truth about rabbits. Rabbits don’t have a lot of rabbits because they have big litters. Rabbits have a lot of rabbits because they breed frequently.
You’ll notice that, unlike the other popularized definitions, I focus on a new variable, “N”, the number of days it takes for your viral cycle to complete. I do this for a simple reason: cycle time matters. The path to success is typically the combination of a high branching factor combined with a fast cycle time. If you don’t think deeply about the channels you are using for viral distribution, you risk prioritizing the wrong features.
How Do You Pick the Right Cycle Time?
Once a growth team digs into the numbers, they quickly realize that there is no one “cycle time”. So what number do you pick for analysis?
There is no right answer, but in general, you tend to find in the data that there is a breakpoint in the data where a vast majority of all viral events that are going to complete are going to complete. For example, maybe with a viral email you’d see most responses happen in 24 hours, with 90% of total responses happening within 3 days. If that’s the case, picking 3 days might be the right cycle time for your feature. Once you pick a cycle time, the conversion rate gets built into your projections.
Cycle Time Matters
If you are already focused on the new user experience, distribution and virality, well then kudos to you and team. Too many consumer products to this day spend too little time focused on these problems.
But if you want to see clear, demonstrable progress from your growth team, make sure you include cycle time in your thinking about what viral features will be most effective for your product.
Now go out and make a lot of rabbits.