We discussed that problem a few times between the current batch of nReduce startups, more specifically there are a few social network projects (or projects with social network features) being developed and nowadays you'll have to strive for outright domination in order to have a chance with a concept like that. This is a tough proposition in a market where the willingness of users to sign up for (and invest time into) walled gardens is becoming very limited for understandable reasons. Which is sad, because I think there are still some great innovations to be made in this sector and the big players have no incentive to do so.
In our case, a few of us are probably going to try and solve this problem by having a federated model that allows us to have one large userbase of people that can interact across servers. This is where my own project fits in, so I'm very excited if we can make some inroads.
As a startup working with this exact challenge, I enjoyed this post. Emphasis on the value prop on both sides is critical, and evangelizing users is your best hope.
Paul DeJoe from Ecquire watches his referrals metrics closely and uses those to judge user evangelism. If a user will recommend you to someone else purely out of the positive experience they believe that person will get, then you've got something.
Sometimes not all chickens or eggs are equal and its not just the amount of each but the specific identity of which chicken or which egg... If that makes any sense.
Not sure if a sports example will fly on HN but here goes:
So, let's say you wanted to create a platform/network for High-School athletes to connect with and get recruited by college coaches. The Athlete can post highlight video, stats, info, etc to gain exposure. Coaches can view these, interact with and find athletes that usually would not be on their radar.
With that as an example you can see where the “general” network externalities model assumptions fail. (where simply the number of and not the individual identities of other adopters matter)
In recruiting there are the athletes who are actively recruited and need not compile their own game film and make/post highlights because coaches have long had them on their radar and will acquire these things on their own. They do not need the network as much as lesser athletes and may actually be hurt by it.
Then there are the majority of high school athletes that may have the talent to play at the next level, be it Div I or III, or even at non scholarship offering institutions.
For these athletes an “Identity-specific” network externalities model is more congruent. In this model each user's benefit is a function of the identities of those who adopt. A lesser athlete's benefit derived from the network would be positive/increasing when an athlete of higher status adopts the network, and negative when someone of lower status adopts.
Then when you to the sport-specific recruiting differences, football/basketball vs lacrosse/soft-ball,.also hockey and baseball being a whole different animal with the inclusion of pro team scouting into the high school athletic experience.
Anyways, just a thought, another layer to contemplate