Lance Wills' Blog – "Internet Lifestyle Trends" (subscribe)!

*Note: It is important to clarify that we are looking at the value of a Fb user from the perspective of Fb, not from the perspective of a Fb advertiser, marketer or user.  However, the methodologies used herein may also produce extremely valuable insights for the marketer using Facebook, and provide an invaluable starting point for your own user lifetime value calculations.

Strategy And Value

Prior to building a model for determining value, an Analyst must know the short and long-term strategy, and correspondingly what exactly carries value for Facebook.  For the sake of continuing our analysis here, I will assume Fb is pursuing a mixed strategy of long-term content and user expansion along with current advertising revenue growth.

1-Year Value Of A Facebook User

To build a model to determine 1-year and lifetime value per user, we will take a granular approach.

First, we will define the model used to determine the value for each individual Fb user, and then adjust this model to account for user engagement archetypes, demographic groups, and finally index this model to user country.  The baseline model will be for users located in the US.

Furthermore, we will use KPIs (Key Performance Indicators) that Fb is most likely already measuring.

The 1-year value per Fb user, can be modeled by first making a few assumptions about what value exists for Fb for each user, and then applying empirical user data to create the actual model.

First, value is hypothesized to be the SUM of several distinct components:

1. Value of the user’s personal network
2. Value of the user’s total time on the site (AKA content interaction)
3. Value of advertising

Each of these can be broken down into further sub-categories:

1. Value of the user’s personal network is a function of:

• User demographic info (income, location, education, age, privacy settings, etc.)
• Total number of Fb Friends
• Number of friends that this user invited to Fb via email (i.e. who were not Fb members prior)
• Number of in-bound friend requests within Fb for this user
• Total number of Groups, Applications, (Product, etc.) Pages
• Application invitations sent out
• Devices configured for regular Fb access (mobile, PC, netbook, etc.)

2. Value of the user’s total time on the site is a function of:

• Time spent CREATING original content (Broken down by media type: pictures, video, products, etc…)
• Time spent CONTRIBUTING to existing content, such as comments (where user may or may not have originated content)
• Time spent VIEWING content (Broken down by media: pictures, video, products, etc…)
• Time spent PARTICIPATING in apps (Games such as Farmville, etc.)

3. Value of advertising is a function of:

• Ad units shown
• Ads served (impressions)
• Ads clicked (ad revenue per user KPI)

Next, we will acknowledge that user engagement varies between every user on Fb, and we will segment all users into three user engagement archetypes. Using empirical data, we could easily define a percentage to each segment corresponding to the percentage of content that each group produces.  For the sake of brevity (since we don’t have that data), we will use a 90%/9%/1% distribution and assign it appropriately, namely:

• Spectators – (90%) those who use the site, but do not contribute, or very rarely contribute
• Contributors – (9%) users who contribute on occasion
• Advocates – (1%) users who contribute the majority of all content on Fb

Each user should then be assigned a value multiplier, or index, based upon their user engagement archetype (e.g. user’s who contribute more content are indexed at a higher value).

Thirdly, we may also elect to dial down value to the user demographic group if empirical data revealed consistent statistical significance:

• Gender : Male, Female
• Age : 0-17, 18-24, 25-34, 35-54, 55+, unknown

Each demographic group could then be assigned value multiplier, or index.  Gender and age would not be handled separately, but would be grouped, such as: (Males, 0-17), (Females, 0-17), (Males, 18-24), (Females, 18-24), (Males, 25-34), (Females, 25-34), etc…

For example, the (Female, 18-24) demographic might have an index value of 1.20.

Lastly, we will index each user’s value based upon their country.  Since our baseline model is for users located within the US, our index for the US is equal to one.  We could quickly derive this index for other countries from empirical data based upon revenue generated from that country divided by number of total users.  However, a more accurate model for each country could be obtained using the same granular approach detailed above; segmenting by user engagement archetype, demographic, etc.

The resulting model for determining 1-year value of a Facebook user is:

User Value = (Country Index) * (Demographic Index) * (User Engagement Archetype) * SUM (Value of the user’s personal network + Value of the user’s total time on the site + Value of advertising)

Lifetime Value Of A Facebook User

Using empirical data over the lifetime of a Fb user we could easily build a model to determine lifetime value.  However, because Fb is a young company and life cycle data may not exist over the entire life cycle of a user, we must make some assumptions in building our model.

The lifetime value of a user will be calculated based on the following criteria:

1. The (first) 1-year value of a Fb user
2. User Engagement Archetype (Spectator, Contributor, Advocate)
3. Life cycle curve fitting

First, we would use empirical data and existing external life cycle data from comparable industries to define curves that best represent user engagement life cycle. Parallels between social networking user engagement life cycle curves and other, similar phenomena would be used.

We would segment and assign each user to one of three user engagement life cycle curves.  Correspondingly, we would tie each one of these curves to our pre-defined User Engagement Archetype categories (this may or may not hold to be valid and may require the creation of a new “User Engagement Life Cycle” category unrelated to the “User Engagement Archetype”).  For example, we are assuming the “Advocates” user engagement archetype would have the longest user engagement life cycle, since they are the group that most frequently generates content, we are also assuming that they will be engaged with the site the longest.

This curve would then produce a polynomial equation that would be used along with the 1-year value (independent variable) to calculate the lifetime value of a Facebook user (dependent variable).

This model is a simplified approach to a very complicated and nebulous challenge; we are committing all users to belong to one of three user engagement archetypes based upon their activity within their first year.  Then we are calculating the lifetime value of this user by plugging the value of the user’s first year of activity in to the projected engagement life cycle curve equation.

However, due to the simplicity of this model we are able to determine the lifetime value of a user on Facebook very quickly, and with only empirical data from the first year (or first several months).  We could easily modify it to include advanced parameters and increase accuracy in the future.  Furthermore, as we accumulate more empirical data and recognize life cycle trends in user engagement, we could very easily modify our approach and look at other KPIs to improve statistical relevance.

Closing Remarks

In closing, there are some factors that I would like to address.

First, I initially wanted to include some of the revenue generated downstream from a user by his/her social network that he/she originated (e.g. ad clicks by a user that our target user invited onto the Fb site).  However, at this stage, I found it to overly complicate the model, as it would involve fractional revenue added to the value of our target user, with the balance of this revenue added to the value of the “friend’s” value.  Furthermore, what about the value generated by the “friend’s” of the “friend?”  As you can see, this MLM approach could become quite complicated very quickly, but could be integrated if deemed appropriate.

Value on social networks is still amorphous.  Revenue generated from ad clicks is obvious, but how about the value of a heavily influential user?  If everyone wants to be a user’s friend then some value exists there, and should be integrated, which I have done.

In conclusion, determining 1-year and lifetime value of a Facebook user is a new and complicated challenge.  I believe that I have created an approach that is both a credible starting point and a powerful calculus capable of satisfying advanced needs, especially given the limited availability of data.