[Note: This is Part 2 of our 4 part series of the Kontagent Social Gaming Workshop Summary, here is:  Part 1, and Part 3]

Paid User Acquisition

When the Facebook platform was still new and there was a land grab to obtain users, paid user acquisition was quite common.  Since then, developers have learned much more about how to grow application using viral channels to keep the cost of user acquisition very low.  This however, doesn’t mean there isn’t a place for paid user acquisition.  Paid user acquisition can still pay very high dividends if leveraged correctly.

Should I pay for users or should I try to grow my application virally?

• Never pay a significant amount for users before testing your application with live users.  You should first ensure the virality of your application is greater than 1 (meaning the application will continue to grow on it’s own) before throwing a lot of money at driving installs, or else you end up wasting money.  If you have Kontagent installed, you get a measurement of virality right on the dashboard
• Understand how much users are worth before paying for users.  There are a number of ways to calculate the “value” of a user, or the LTV (LifeTime Value).  The simplest way is to find out on average how much a user generates over the lifetime of the user using the application, whether that’s through ads, offers or direct payments.  However, this doesn’t leverage the viral factor which is available to us.  Using the viral factor, the real value of the user is not how much the user is worth, but what’s the value of the network of users resulting from this user inviting other users.  This is called the Lifetime Network Value of a user (we’ll be doing another blog post that dives deeper into determining user value)

How much should you be willing to spend to acquire these users?

• This is closely tied to the answer above about the value of a user.  If you understand the value of a user (the Lifetime Network Value) and how much it costs to acquire a user through a paid acquisition channel, the answer is then pretty simple.  If the cost to acquire user is greater than the predicted Lifetime Network Value of a user, then it makes sense to pay for the acquisition of a user.

How well do cross-promotions work?

• Cross promotion can be a highly cost efficient way to seed traffic to a new application, or just to grow an application if you have other application to promote from or you are cross-promoting from other applications you are working with.  What you must look out for is there are a large number of common users between the 2 applications.  A couple of developers shared their experience about doing cross promotions with the discussion group and determined that there is typically high-churn when doing cross-promotions.  Even if 2 applications have the same gameplay, the genre can also make a significant difference in determining whether the cross-promotions will work.  You should run a small test to see if cross-promotions result in returning users from the application that is being promoted.

How do you identify the key characteristics of users who are likely to pay to play?

• This is a very interesting question.  Simply taking a look at the data and correlating user behavioral metrics to tendency to monetize would answer this question, but this is something we haven’t done yet.  If there is anyone who is interested in providing us with data to determine this, please send me an email (info(at)kontagent.c0m)

 

Viral User Acquisition

Viral optimization has been a hot topic for quite some time now and we will have another article that is just focused on viral tuning using Kontagent, so the topics that were discussed in the workgroup were focused on less discussed topics surrounding viral tuning.

It’s well know that the number of users converted per user from a viral event is a good metrics to look at for viral user acquisition, but what are the right users to look for viral tuning?

• Just observing the number of users acquired per user [ invites sent * conversion rate ] is a great first order metrics, but it is possible to go even deeper and take a look at second order effects.  The question then becomes, what is the quality of users that are being invited in terms of both virality and engagement?  The metrics is then to take a look at the number of users acquired by each user in the second generation and then to take a look at the engagement of the second generation users as well.

How do you optimize the virality of a user in the long-term rather than just the short-term?

• When viral optimization is discussed, the most common metric that’s used is [ number of outbound messages * conversion rate ]which optimizes the number of users converted per event.  The shortcoming of this model is that it’s meant for short-term optimization rather than long-term optimization of virality.  While this optimization is great for spreading video clips or articles virally which are typically one time events (when was the last time you revisited a video to share it with another group of friends?), it’s not the ideal metric for games because games are designed to engage users repeatedly.   This means, it’s important to optimize the users converted per event as well as the number of times a user repeats the viral event.  For example, say the context for an invite event is “invite your top 5 friends to play a game with you”.  In this case, the conversion rate may be high for the event,  but the likelihood of a user coming back to the same event is very low since a user’s top 5 friends doesn’t change very much.  Effectively the lifetime virality of the user is lowered as well.  In contrast, an invite context such as “invite 5 new friends to play a game with” may not net as high a conversion rate per event, but the user may repeatedly come back and so the overall lifetime virality is much higher.  In summary, you should be optimizing [ number of outbound messages * conversion rate * number of times the user revisits the viral event ] which can only be done if history is kept on each user.

How does the strength of friendship affect viral invite conversion between user X (sender) and user Y (recipient)?  Is there an actionable metric than can be used to tune viral events?

• When user Y receives in invite from user X  there are three key factors that determines whether or not user Y will accept the invite:  1) what the application is 2) the content of the message 3) who sent the invite.  If we think about it a bit more, it becomes pretty obvious that the stronger the friendship relationship, the more likely you are install the application.  Implicitly, people typically stay in close contact with friends that have common interests which leads to more trusted recommendations.  Another way to think about this question is:  would you more likely try an application recommended by someone you barely know or a close friend that you keep in constant contact with?  From this perspective, the answer is pretty clear, you’re more receptive to close friends.  Now let’s take a look at an invite event.  If we know that stronger friend ties trend toward higher conversion rates, what actionable metric can be used evaluate an invite event?  We’re proposing a new metric here:  [ % of close friends invited/total friends invited ].  We have yet to prove whether or not this metric is actually useful with real data, but we’ll post some results when we’re able to get some empirical evidence.

Tags: metrics, social gaming, summary, workshop