-
Website
http://andrewchenblog.com/ -
Original page
http://andrewchenblog.com/2008/03/05/facebook-viral-marketing-when-and-why-do-apps-jump-the-shark/ -
Subscribe
All Comments -
Community
-
-
Top Commenters
-
Popular Threads
GREAT Post Andrew! I'm currently developing a Facebook app with a partner and one of my chief concerns has been whether the value of the application is sustainable of the long-tail.
This post gave me a few ideas to lessen the sting of user burn off to help re-engage them in the application.
Thanks for taking the time to cover this!
Well done on realizing that the math used by biology and ecology folks applies to Facebook!
I make my entire living on the concepts you outline. I apply those concepts to companies, outlining for them how their business may be struggling due to the issues you describe in your post.
You correctly identified that adoption/acquisition, carrying capacity, and retention drive all dynamics, and do so not just for Facebook, but for all business models.
Well done!
that's awesome, Andrew. Folks have been hinting at the math behind viral distribution on FB, but this is the first time that I've seen it broken down.
Thanks. Great details. Most appreciated.
As someone who built a viral app (pre-Facebook) I primarily nurtured and grew it for many more months than another entrepreneur might simply to make sure if I was going to commit to it that the users were sticking around.
Of course I tried to get as much easy revenue as I could early on (gotta get the low hanging fruit ;), but I didn't invest heavily into it until I confirmed it offered long-term value to the users and the long-term reward would match the long-term effort.
I've really been appreciating you posting these formulas and calculations. Where as in the early days it was fine to throw stuff on the wall and see what sticked, we really like to make sure if we put focus and effort into something it's not just going to stick, but climb up to the ceiling ;>
Thx
Wrote a piece on the same subject a few weeks ago....the similarities in our graphs are uncanny - great minds, eh?
Here was mine:
http://www.broadstuff.com/archives/750-The-Rollercoaster-Dynamics-of-Social-Net-Usage-Traffic-Crash.html
i have nothing to add except screw the math warning-this post kicks ass!
thanks!
Andrew,
Yay! We've used similar models internally at Adonomics. I have a BS in Mathematics, so this stuff makes me happy.
If you're interested in any of the data we have I'd be happy to share with you. We're the only ones with data before the switch to DAU, so we actually have detailed graphs of user growth.
You can email me at jesse[at]adonomics.com if you're interested.
Cheers,
Jesse
CTO, Adonomics
Love this great post.
The logistic curve you refer to is actually also known as a Sigmoid function (aka S-Curve).
On Wikipedia:
http://en.wikipedia.org/wiki/Sigmoid_function
There is also an Excel spreadsheet floating around for the S-Curve here:
http://jcandkimmita.info/jc/2007/04/business/modeling-market-adoption-in-excel-with-a-simplified-s-curve/
I've used this to model pretty closely Facebook application growth of some pretty popular apps.
Best,
Stanley
Well written article, but I respectfully disagree. The hypothesis may not reflect social realities.
Thousands of Facebook members have endorsed termination of widgets that force more friends to install. This push strategy is socially out of date.
Social applications win by pull. As each friend installs, their friends discover the application and installs. Thus, as recommended by Mark Zuckerberg of Facebook, developers should focus on great products, not viral marketing.
Bloggers win via persistent pull. As you write each post that expands the relevancy of your content, more readers install your feed - spreading the word to their friends.
With an application or blog, the result should not be modelled as a one time event producing a return curve. It comes from incremental improvements to the application and content - producing long term growth.
Social networks provide only the tools for potential viral growth.
-Dash
The New Economics of Advertising - http://adEcon101.blogspot.com
Great post -- appreciate the math breakdown.
It's interesting that the first curve you post IS the traditional "hits" model of entertainment sales -- hit movies, hit singles, hit games. This curve dives quickly when it's a one-time content experience: purchase the hit, enjoy the hit, then move on.
With multiplayer games and social networks, we have ongoing streams of fresh content - and the potential to direct viral growth into an ongoing, ever-changing entertainment experience.
SOME app is going to be the WOW of FB - the breakout hit that grows virally and takes over an entire category - not through a spammy invite system, but because it's SO MUCH FUN to play together with your friends and family.
Venturebeat reported record growth on fubar.com on Mar 7, while Alexa already showed the traffic at the jump the shark stage, like your graph.
Great post. Seems to me that this just shows why you need to always be in beta or working on the next version... There are few long-term home-runs in this business - and all you are really doing with great products is lengthening the shark tail, not eliminating it.
Seems to me that a possible strategy with this reality is using your initial product to popularize the second and so on... Overlapping shark-tails means when you jump the shark, the users have somewhere to land. Just a thought.
Andrew,
You've generated an all star discussion forum in the comments! (myself excluded of course).
One refinement I'd suggest based on cohort analysis of subscription businesses is to vary retention churn over time. It is typically not a constant x% per period. Rather you typically get high "infant mortality" over the first few periods, but then settle into a steady state with little loss once you've identified your core users. This data is now several years old so I feel comfortable sharing it, but in AOL's dial up business around 2002 we would average around 6-7% churn per month. But this was an average across multiple cohorts; we would lose 50% of users in the first three months, but by month 12 churn was in the 2% range (which is close to the move/death range of unavoidable churn when dealing with a service tied to an address/phone number). This "mix" problem is particularly important when you're modeling viral growth businesses that have meaningfully different user acquisition rates over time.
Any particular reason why you took this approach instead of applying the Bass model to this? http://en.wikipedia.org/wiki/Bass_diffusion_model
Great post.
Clearly you've mapped the short lifecycle of most Facebook applications at this point. This kind of theory would hold for any sort of engaging application online. The anomalies do exist though, like Top Friends or Compare People, where constant innovation keeps users coming back for more. I imagine this is what we'll see after Facebook shakes down existing applications with the new Profile setup.
I didn't understand some of the maths bit - but understand the implications loud and clear. Thanks for the post!
-DH
Visit my tech blog @ http://techrunch.blogspot.com/
Great post, one thing that struck me is that it doesn't capture the "competitive equilibria" issues that may also occur. Your social network isn't the only one recruiting, so that there is competition for the total carrying capacity. I think there may also be a bandwagon effect, where users depart smaller groups (in the context of a particular demographic/market/segment that you measure with your carrying capacity) for larger ones because they offer more benefits. But I agree with your basic modeling approach.
Hi Andrew,
I did a talk a while back on your findings to the Facebook Developer Garage in London, which I help organise. After posting about my talk ended having an discussion with a reader over some inputs which create a saw-tooth in the results, you might find the conversation interesting - http://www.joshuamarch.co.uk/2008/04/jumping-shark.html
Cheers,
Josh
Definitely agree with you Andrew, retention is really important when developing a new Facebook app, or any app for that matter. Great math, great stats, and great post.
Hey, i just found out about this theory on Google I/O. I will defintlly look for more of syour developments. I have made a post about organic growth vs viral growth on my blog. Is acctualy in spanish, so you might find it interesting:)
http://www.lawebdejuan.com.ar/2008/07/crecimiento-viral-vs-crecimiento.html
Awesome work!
Hi Andrew!
I played a bit with your excel-sheet and your equation and explored that your results in excel doesn't match with your posted equations. You never used something like ^t.
In excel you use something like
u(t)=u(t-1)*((1-u(t-1)/carring_capacity)*conv*i)+u(t-1)
Where is the mistake?
This is good methad and good story.
NevaehEastkbff
----------------
Wow, check out this site called http://www.fluc.com
. Free SMS and free mobile ads!! Its fantastic