What viral coefficient actually is?
Product’s viral coefficient is the answer to the question “How many additional users one new user will bring organically ?” If for example Facebook found that each new user will usually bring 5 additional users by inviting his friends through email API, and 1.2 which will join because of this person is telling his friends about how wonderful Facebook is, Facebook can conclude that their product viral coefficient is 6.2.
Calculating viral coefficient
There are two methods to calculate viral coefficient. One is quite accurate, cookie tracking based but usually not feasible. The other one is based on high level estimations, but actually can work for ANY website or product. 
- Calculating Viral Coefficient
Method A:
Imagine only people who has invites can become your users. Imagine only existing users issue invites. Imagine that you can attribute each new user to the inviter at the individual level. Now stop imagine and look at how gmail, Quora and Pinterest started. These website (Any many others) started their service as a closed, invites only party. Besides the fact it makes the newcomers feeling prestigious, it allows these website to calculate viral coefficient in a very accurate way. Each invitation gets a unique tracking URL, hence each newcomer is necessarily coming through a unique tracking URL. It’s enough in order to calculate how many new users each existing user brought. It’s an excellent way to calculate viral coefficient. The only problem is that this method can be used only at early stages of product’s life. Usually products who like to grow, must open up their gates.
Method B:
Being realistic, most websites and most brands will serve users regardless to whether they received an invite from existing user or not. This very different situation requires a very different approach to viral coefficient calculation. Using method B, we will look at the total number of users we have and split them into two buckets. In the first bucket, we will place all users we had to bring proactively. It’s obviously includes paid efforts such as PPC or media-buy, and often includes SEO and referrals traffic if these not happened organically. The 2nd bucket will include all the rest – users that arrived virally! The proportion between the two should give us the viral coefficient, but WAIT!
One important correction will make your calculation even more accurate and much more prudent. Cookie tracking is reliable to a limited degree. As a rule of thumb 20% events are not tracked properly by the cookie. What it means is that your first bucket is actually 20% bigger then what you could measure. The second bucket is necessarily smaller than was calculated initially.
I find method B to be exciting and very pragmatic way for viral coefficient calculation. What do you think? I’m really curious to know!
P.S
When calculation viral coefficient using both methods, it’s critical to remember that it’s a variable that changes throughout the product’s life. Facebook viral coefficient today is very much different than what it used to be 5 years ago. This comment puts viral coefficient in the right perspective. Viral coefficient is a very good indicator for product’s virality, but it’s always based on historical data. Historical data often does not project the future.
I just happened upon this approach trying to figure a way to calculate viral coefficient with limited data on invites, and I’m drawn to the “Method B” you describe, where K = viral signups / regular signups. I’m curious about viral cycle time though, and how that impacts the measurement.
If you measure over one viral cycle, such that viral signups = regular signups * viral coefficient, then it works out, but let’s say for example you’re measuring over four weeks, and your viral cycle is only a week. For sake of example, we’ll say each week you get 100 regular signups and our viral coefficient is 2.
Starting point: (100 new users)
Week 1: 100 regular, 200 viral (300 new users)
Week 2: 100 regular, 600 viral (700 new users)
Week 3: 100 regular, 1400 viral (1500 new users)
Week 4: 100 regular, 3000 viral
So in this case, over 4 weeks you have 5700 new users, 400 regular and 5300 viral, but the viral coefficient is only 2, not >1000 as measured by Method B. Have you considered a way to measure over multiple cycles?
Well, that’s a very good question, and that’s obviously the biggest disadvantage of using method B. While I don’t have a clear answer for you, I can tell you that usually, the solution revolves around generating a controlled traffic peak (“Regular”, often paid) and than to analyze the secondary (“viral”) traffic pattern. Understanding the Viral Coefficient cycle time is the missing piece, and this method can help to estimate it better. Hope it helps
Pingback: Quora