Startup Over-Confidence of Network Effects

Network effects have been on my mind lately–a lot–and since Fred referenced the term yesterday I made a few comments and did some core research to validate/invalidate my assumptions and gut.

I  feel there’s lack of understanding about the dynamics and efficacy of network effects. Typically we refer to Metcalfe’s law and talk about a pure environment, where every node that’s added to the network is universally and always available. 

When we apply this to human networks, the environment is not so pure and not reliable. We need to consider  half-life, diminishing returns, dead stops, time limits of each human node, etc.

When applied to web-connected devices (nodes), flow is largely unimpeded, with messages/signals reaching the intended destination, and the response is highly likely (even taking into account the rare device failure, like a switch going down, or your wifi router, etc). 

But networks of people have limits because people have limits like time, accessibility, interest, belief, trust, need, ambition, acceptance, etc. 

So when we apply the term network effect to networks of people (social networks, trade networks, interest networks), we have to consider that they are not always on, and not reliable, like a network switch or device etc. 

Those human elements are variable in intensity and availability, somewhat unpredictable, and different across people. 

And the currently favored delivery networks (FB, Twitter, LinkedIn) present your messages in temporarily visible ways (as opposed to email, sms, voice) because of the use of streams instead of static messages designed to go from you to someone else who will most likely read anything in her inbox (unless it’s poor Fred, inundated with email). 

But let’s just say the only network is the phone network, and you want your message to go from A to H, and the connection is assumed to be constantly available; the chance of your message reaching H without a direct call to H is something like, oh, well it’s low (that would require math), given the human variables that leave gaps in the message chain. 

I’ll reference Leslyn’s startup (disclaimer: we have not discussed this topic so this is not a reflection of Leslyn’s thoughts or model necessarily). 

When thinking in broad, general terms about a base of people (suppose there are 1 million people who would pin something on a map and share it with their 150 Facebook friends, thus our potential reach is 150 million people), keep in mind that 

  1. getting to 1 million committed, regular users is really tough (and a different topic),
  1. not all 150 FB friends will get the message, even if each person sends a direct message (streaming content is only seen by people viewing the stream, or who seek out and browse your stream after the fact)
  1. even if 30 get the message, perhaps only 3 will click through to see what was shared (limited by time, interest, relevance, need, etc)
  1. of the 3 who click, it would be a miracle if 1/3 of them become a new committed user. It’s more likely 1/40, because our social networks are largely aggregates of interest networks, not a single network of people with the same interests. 
  • Reach is not 100%; measure reach and apply it to your models. 
  • Expressed interest (click-through) is not 100%; measure expressed interest and apply it to your models. 
  • Adoption is not 100%; measure adoption and apply it to your models. 
  • Production (pinning and sharing) by your committed, active people is not 100%; measure the sharing of pinned stuff and apply it to the model.

  • And make sure you reward that behavior and remind them to come back and enjoy those rewards more (create a desire engine). 

I’m not an economist, researcher, mathematician, or focused marketing guru–I’m just observing, learning from others, and trying to understand ways to get past the gaps and limitations of human networks. 

I’m not an economist, researcher, mathematician, or focused marketing guru–I’m just observing, learning from others, and trying to understand ways to get past the gaps and limitations of human networks. 

So be realistic when thinking about networks of people. 
What you will end up with, in simple terms, are percentages multiplied by percentages, which produce much lower engagement results than the perfect network model. 

Your equation is R*E*A*P (this was totally accidental, btw but I like it). If R=20%, E=10%, A=10%, and P=50%, your resulting network effect growth factor per period looks like this : .2*.1*.1*.5 = .001. 

If you have 10,000 users, and this is your monthly growth, your increase is 10,000*.001, or 10. 

That’s a pretty weak number. 

So your work then is to optimize at each stage, and add in other, more dependable things like email, SMS, phone calls, letters, visits (depends on your business of course), but not so much that you discourage people from using your system. 

And you can optimize the frequency/reduce the amount of time your people produce and share. 

If you have thoughts or articles to share about these issues, please comment below. If you want to read real research on this with a less linear approach, check this out.
  • Leave a Comment

    Your email address will not be published. Required fields are marked *