Oct 5

Tim O'Reilly

Tim O'Reilly

Good News, Bad News about Facebook Application Market: Long Tail Rules

My team at O'Reilly Research has been crunching the numbers on the rise of Facebook as application platform, and we've released a new report today, entitled simply The Facebook Application Platform.

The good news has already been widely disseminated: there are nearly 5000 Facebook applications, and the top applications have tens of millions of installs and millions of active users. The bad news, alas, is in our report: 87% of the usage goes to only 84 applications! Only 45 applications have more than 100,000 active users. This is a long tail marketplace with a vengeance -- but unfortunately, the economic models (for developers at least, though not for Facebook itself) all rely on getting into the very short head. Here's the distribution of active users among the top 200 developers. (Some developers have multiple popular applications.) As you can see, the drop-off is extremely steep:

Facebook long tail

This doesn't mean that Facebook won't become an important platform for developers, just that a throwaway Facebook app is not the ticket to quick riches. Embracing the Facebook opportunity requires more than just optimism.

The report includes details on the top suppliers (developers with multiple applications), the average adoption rate by category (some categories are more competitive than others), and "historical" growth rates for the top 200 most popular applications. (I say "historical" in quotes, because the history is so short. This is a marketplace where anything can happen. As a result, anyone purchasing the report will receive access to two downstream updates.) The report also includes interviews with Max Levchin of Slide, the top supplier of Facebook apps, and Ali Partovi of iLike, also one of the top suppliers, and thoughts on Facebook as a social network operating system, plus Facebook marketing advice from Dave McClure and overall market perspectives from Niall Kennedy.

While the report focuses on the market for current Facebook applications, it seems to me that the future opportunity is less in Facebook applications per se, and more in the development of applications that use the social graph embodied in Facebook for entirely new purposes.

I'll be presenting data from the report, and my thoughts on the next generation of social graph applications, at Dave McClure's Graphing Social Patterns conference next week. The conference will be held in San Jose, CA from October 7th-9th. Main conference sessions are Monday 10/8 and Tuesday 10/9; an optional pre-conference workshop is Sunday, 10/7.

If you're working on a Facebook application (or other social networking application!), you might also want to come to Appnite at GSP, with a hackathon that kicks off Sunday evening 10/7 , and a live AppNite Demo Contest session on Monday evening 10/8.

Update: A number of readers have chastised me for reversing the sense of the "long tail" in this piece. They're right. It should correctly be called "the short head rules." The significant news here is that the tail doesn't show lots of volume, not that it does, which was Chris Anderson's original point about the long tail.

tags: web 2.0  | comments: 43   | Sphere It

Previous  |  Next

1 TrackBacks

TrackBack URL for this entry:

Tim O'Reilly and team have done some interesting research into the rise of Facebook as application platform. They released report entitled The Facebook Application Platform, which we've been looking at in the virtual offices of Read/WriteWeb. One of th... Read More

Comments: 43

  Claus [10.05.07 04:39 PM]

Long tail with a vengeance? Wasn't the point of the Long Tail piece and later the book exactly that it does not look like this?
The Long Tail argument claims that the tail is fat, that there's a realistic useful demand all the way out, but what you're pointing at indicates pure winner takes all.

  Tim O'Reilly [10.05.07 04:59 PM]

Claus, not really. The point of the Long Tail book was that the demand out on the tail AGGREGATES to a significant number (and can make it fatter) but not that demand is significant for any player on the tail. I wrote about that in a blog post last year called The Economics of Disaggregation: "long tail businesses disproportionately benefit the aggregator. While they create new opportunities for content providers "down the tail" who might not otherwise have been noticed, they create even greater collective benefits for the Amazon, the Google, the Netflix, who hosts the entire collection, the dog who wags the tail."

It is true that effective online marketplaces do fatten the tail. (See my post on long tail effects in Safari and Google book search.) By that measure, Facebook is not yet an effective long tail marketplace.

Of course, it's very young, and over time, there may be a more even distribution. But right now, it does look very much like an extreme winner takes all marketplace. That actually makes sense if you think about it, as developers are competing for a very limited slice of the user's "attention real estate."

  Claus [10.05.07 05:21 PM]

I think you're stating my point - but maybe I made it badly.

  tom s. [10.05.07 05:24 PM]

I agree that our host is restating Claus's point. The data you quote suggest that even the aggregate demand in the tail is weak, whereas the point of the book (which is wrong, in my opinion) was that aggregate demand is large - even enough to outweigh the head. So facebook applications do not form a long tail market.

  Tim O'Reilly [10.05.07 07:04 PM]

Claus and Tom --

You guys are right. I'm misusing the term. Chris' point was that aggregate demand out on the tail was large. And in fact, here we see that it's relatively small.

My point was that most people will find themselves out on the tail, and that it's a skinnier tail than they might expect.

But I'm not sure that's entirely inconsistent with other long tail markets. It's just relatively short. If Amazon sales drop off to a few units a week after a hundred thousand titles, there are still millions more titles out in the tail, and that does add up. But when there are only 5000 facebook apps, the tail isn't long enough for the aggregate demand to rival a significant fraction of the head. As facebook grows, the aggregate tail will too.

  Crash [10.05.07 07:40 PM]

Perhaps a better economist catchword here is the "tournament model": in a field of a hundred competitors, only the top three make money, to some part subsidized by all the rest.

  Thomas Lord [10.05.07 08:04 PM]

I interpret the data this way:

As an application platform, Facebook is apparently not successful. Five thousand first applications of the platform -- applications that make it all the way to public release -- and fewer than 100 are of any general utility. That suggests a too high barrier (the cost of creating a new app) for too low a return.

Compare statistics to the early days of past, proved platforms like the original unix. By the time unix had about 5000 public programs in that community, many many of them were in heavy use by many customers. (And, the virtuous long tail was starting to appear.)

What's the difference? What makes one "platform" so productive, like unix, and another so anemic, like facebook?

The answer is that Facebook is barely a platform at all: it is a client side library and a small set of web services. The web services mainly focus on the newly invented social graph. The primitive operations on a social graph don't rise close to the generality of primitive operations like storing files, running processses, and making pipes. Moreover, nobody has any really convincing answer to the question of how users benefit from the creation of a social graph and its exposure to application writers in this way -- it seems a fairly random experiment if the goal is to transfer value to users.

So, facebook is very fancy and cool and does what it does but it's not a platform -- just a browser library and some AJAXy web services of dubious value. The mythical "platform" people are waiting for lies in the future, behind a different technological underpinning.

(The other angle that's popular on Facebook these days is that it is exploiting hoi poloi's new found fascination with "on-line presence". One has to wonder if that new found fascination doesn't exist only because efforts to incite it have gotten the most investment. It's funny how just about no upper class person would publicize their on-line presence that way at all, except when they were trying to prove the point that they're willing to.)


  Michael Jones [10.05.07 09:07 PM]

I'd have to disagree with one point of Thomas' above...if you take the grand total of all things Linux (or for that matter, integrate shareware/freeware into the Windows marketplace) you get a curve quite similar to this. Most Linux apps are like most freeware apps - of questionable quality and limited audience. There's not much unique here in the grand scheme.

I suspect a map of Google Map mashups would look similar - we just aren't aware of those on the far reaches of the tail.

What makes Facebook problematic is that all apps - especially the obnoxious ones - are very much in your face, popularity or no. I was referred here by a friend who was pissed by my cornucopia of apps on my wall (including such gems as the Lesbian Dectector, which clearly doesn't work - I'm way more than -5% lesbian.)

The balance may just be that I get reamed out for posting such crap on my profile, which makes me consider the value of what I include.

One of the great benefits of Facebook is that it's not as chaotically random or full of spam as MySpace or the like. A danger to Facebook apps is that it might make FB less "sticky" through introduction of chaos and random invites to become vampires. It's cute the first time, but gets stupid very quickly.

What is more, by opening the API up, Facebook does risk losing all control over the look and feel and interaction habits of the space. The API is is calculated and potentially valuable risk, but it is a risk.

  Deepak [10.05.07 11:01 PM]

Very interesting post and comment thread. I must admit that I would not enjoy using Facebooks but for the apps. Let's forget about the long tail for a second. My interpretation of the data are that those that think a Facebook app = great viral marketing might be shortsighted. It has to be the right kind of app.

However, many of the apps on Facebook are not necessarily intended to be marketing vehicles. Some are lighthearted, other rather useful Lets take an app like Bookshare for example. It's just meant to be an app for people to find out what books others like. To me that's a great use of the kind of social platform Facebook provides. It doesn't really matter if only a 50,000 people are using it, since that would probably be sufficient for most hobby developers and for the user. In fact, if just my friends used it, it would probably be fine for me. I suspect that's what you are alluding to when you say "new purposes"

The social graph/network manifests itself in another way. You are much more likely to install an app your friends install, so an app like Dopplr or Bookshare is likely to be more useful to my social network than HotorNot, which almost none of my contacts use.

  Ryan Miller [10.05.07 11:18 PM]

The graph above shows popularity of apps on the Y-axis, which is interesting. Get REVENUE on the Y-axis and it will be a LOT more interesting. And I wouldn't be surprised if the rank order changes a bit.

  Mark Mayhew [10.05.07 11:43 PM]

good post, although I know more than a few Facebook Fanboys that are not gonna like it!

  Sally [10.06.07 06:38 AM]

Interesting analysis. The only way someone could, in theory, make money by not being in "fat head" would be roll up a significant number of small facebook apps in order to aggregate enough of an audience to make it popular. I believe in terms of creating a network, this is what the Facebook Ad apps are trying to do.

However, I think the whole thing is clearly a losing proposition for developers. There is only one party that will actually succeed from Facebook as a business in any real way, and that's Facebook.

  Anonymous [10.06.07 09:18 AM]


  danny [10.06.07 09:40 AM]

It would be nice to see a chart of Windows consumer apps by comparison. I'd imagine a similar curve, with MSFT, Adobe, Quicken, Yahoo, AOL, Google, couple media players and IM clients at the top, and most others far behind.

And it would be nice to see a chart of TV channels or magazine subscriptions by comparison. The tail would likely be longer (especially if you look by brand rather than parent company).

These charts would then properly summarize the Facebook quandary (software platform vs. media platform)...

  casey kazan [10.06.07 10:37 AM]

We have two apps on Facebook in the Education Directory, each with 500 users (with fluctuating daily usage). We've been live there about a month and are growing slowly but steadily day by day. We view our FB presence as a cumulative, long-term adoption process vs the hype of the "short-head." And that's pretty cool!

  D. Andersen [10.06.07 11:41 AM]

"Marketplace?" Um, there's no exchange of goods, just people using free apps, that's no marketplace.

Were any of you around for .com 1.0 or do you have collective amnesia? Most of these social networking sites, Facebook included, will never make significant revenue from end-users.

  Claus [10.06.07 11:44 AM]

Another point worth mentioning: Isn't this curve what you would expect relative to the long tails for culture, i.e. books, CDs, movies etc?
Since so many facebook apps are social in nature, the value of facebook apps (to users) grow as adoption grows - classic network economy externalities.

For the cultural products it is much less obvious how the value of a book or a CD grows, for the individual reader, as the fanbase grows. Certainly, social value is not the only or even the dominant value in a lot of the cultural goods.

  David Geller [10.06.07 12:11 PM]

Great post! It seems to support some comments I heard presented at a talk the other evening where Snapvine's Joe Heitzeberg and Picnik's Jonathan Sposato spoke. They both suggested, too, that Facebook's recent change in the level of viral exposure a single application can make has contributed to the slowing of some application's popularity.

One of the nice things about the FB model, though, like the web, is that the barrier to entry is quite small. So, while a company may produce something that, in Casey's case, only has 500 users, the cost of getting to that point can be relatively low. Low cost and short development time equates, somewhat, to low risk.

No one knows for sure where Facebook will go. It rose very, very quickly. That also suggests it's entirely possible (and likely) we'll see competing social platforms develop and reach similar stratospheric growth patterns.

  Alexander van Elsas [10.06.07 01:18 PM]

Tim, really interesting research. I won't try to make any long tail comparison, as others have done that quite well. But to me the figures show that success on Facebook is just like in the rest of the Internet market. If you don't create value for the user, they won't use your application:

I guess so far about 84 companies got it about right, and even less (45) are doing ok. Which is to be expected. too many start-ups are trying to live off the Facebook success without actually thinking about user value. But hey, investors like it right? I'm off now writing yet anther useless Facebook app :-)

  schip [10.06.07 01:42 PM]

Assuming I've understood some of the dribs and drabs I've encountered over the past decade...
A "fat tail" is technically a larger distribution of 'stuff' in the farther reaches than would be expected from a gaussian. This behavior often results from a power-law, and you can tell if you have one by plotting your data on a log-log scale and looking for a significantly long straight line. E.g., the distribution of links on the web follows a power law and can be generated with a preferential-attachment algorithm. I can go dig up some papers if needed, or you may find useful info by crusing, Newman and Farmer come to mind.
Since the X axis of the graph has no scale it's hard to tell if it's a fat or skinny tail. It might be useful to plot a gaussian with about the same stddev over it for comparison.
Nevertheless, fat-tails are not-quite orthogonal to the question of money to be made. Tail aggregators such as Amazon probably have a better chance with a fat one, but the more interesting detail is likely to be the stddev and the point at which the non-aggregators put their cutoff.

  Scott Rafer [10.06.07 01:48 PM]

A significant number of the multi-application vendors who constitute the shoulder publish apps under pseudonyms. The shoulder is much healthier than you represent.

  Hugh Toppe [10.07.07 12:05 AM]

Your point is a bit of stretch, asserting a generalization based on apps that, in the vast majority of cases were built by individuals (not even companies) over the course of a few days or a couple of weeks. RockYou and Slide are the few exceptions that prove the rule.

The real impact is yet to come: when apps that are now being built, by teams of three or four full-time experienced developers (rather that part-time midnight programmers) over a period of several months, kick in.

I think that 90% of the "short head" apps, the blockbuster apps on your chart, will lose 90% of their active users 12 months from now, as the novelty wears off. But these will be replaced a new generation of apps that deliver greater long-term value.

Bottom line: Interesting chart, making for fun conversation (like a Facebook app), but it's way too early to extrapolate from this data. It's like looking at a fish-looking embryo in the womb and trying to decide if it will grow to be a human or a horse.

  Tim O'Reilly [10.07.07 08:12 AM]


I agree that it's very early, and that there's lot more to come. But note that that's what I said too: "While the report focuses on the market for current Facebook applications, it seems to me that the future opportunity is less in Facebook applications per se, and more in the development of applications that use the social graph embodied in Facebook for entirely new purposes."

  Jane Quigley [10.07.07 09:16 AM]

I agree with Hugh - What also strikes me is how many applications stop working after a few days, weeks or just have so many consistent issues that I get rid of them. Even the Twitters and the Powncers. So the quality control isn't there yet.

Tim - It's a Wild West approach, but it tests the platform - I believe it'll settle down as established companies see the opportunity and start to play(having learned lessons from those that came before). I'm also excited about different ad models - You have companies like VideoEgg that uses the FB application platform to create an ad network. New opportunities for brands to interact with their audience!

  Adam [10.08.07 11:50 AM]

Deepak I think makes a good point:

("However, many of the apps on Facebook are not necessarily intended to be marketing vehicles. Some are lighthearted, other rather useful...")

We are comparing apples to oranges on Facebook, compared to other long tail aggregators (Amazon, Netflix, etc.) where it's apples to apples. This long tail analysis, with all due respect Tim, is flawed. You cannot just compare a distribution of apps as you would, say, books or music CDs. The very nature of software.

  Jessica Margolin [10.08.07 01:07 PM]

It's a sperm. The tail is necessary for motility, but it's only the head that makes the baby.

  Jessica Margolin [10.08.07 01:16 PM]

You know, when you have too MANY aspiring developers/actors/musicians, it turns into a situation where people just hope to be discovered. So if, as a society, we could be quiet already about the alleged "shortage of developers" and realize that it's a lack of companies that good developers want to be with we might actually make some progress.

Facebook is providing a venue for talented, creative developers to communicate to the world what talented and creative developers could possibly do. It's also got a lot of drek.

  Allen Stern [10.08.07 02:02 PM]

There are several concerns I have with this report - you can read them all here:

But suffice it to say, there are easy ways to manipulate those numbers and in fact Mike Arrington wrote a lengthy black hat post about 2 mos ago which included the top 3 as the biggest offenders!

  Anonymous [10.10.07 09:17 AM]

> You guys are right. I'm misusing the term. Chris'
> point was that aggregate demand out on the tail
> was large. And in fact, here we see that it's
> relatively small.

Umm, no it's not. This is just the top 50, and my eyeballs tell me the top 5 and the next 45 are about even, so presumably the addition of the other 4550 does outweigh the head, if not substantially.

  Tim O'Reilly [10.10.07 09:47 AM]

Anonymous -- I'm afraid you're wrong. The dropoff is SO steep that there's no way that the tail outweighs the head. I'll do a followup post with more detail.

  Amina Bayou [10.10.07 12:39 PM]

I'm with iLike, haven't read the report, but wanted to comment briefly ~ a commonality between the top installed apps. is 1) they have all been in the works for many years, and 2) most got in first.

Most of our explosive growth (i.e. 350,000 new users a day) was within the first couple weeks which lead to sustainable viral growth. Only 50 companies or so participated in the FB platform launch and these companies had to have the foresight (as well as resources ~ quick turnaround)to realize how this could ultimately redefine social networking. iLike was originally, which has an almost 10 year history in the .com music space. We are thankful that users have embraced our service, but it's not purely coincidental. Timing and many years of technology development were on our side.

  sumit gupta [10.11.07 07:13 PM]

Great discussion. but isnt it too soon to measure the long tail for this nascent internet market. Its like trying to analyze internet web 1.0 economy in 1997. Not because critical mass isnt there yet, but its because the engagement is still so different for different age groups.

Clearly early application developers have had a head start and because network effects is the gas pedal on which this market breathes, they are galloping while the laggards remain envious and wishful. But the success in engagement they have had is with stupid (and funny) apps which teens, students. Clearly the older age groups have started to participate, but is it to observe or is it to participate. Fred makes the point in his musings that participation drops off like a waterfall once the years weather off. Facebook is still predominantly a 15 to 26 year old service. The usage drops off dramatically once people get above 30 years old.

Research from a recent user study (email me the link if someone finds it, I cant find it) shows that more than half of Facebook users are not currently enrolled in a university or college and that the site’s fastest-growing demo is the 25+ age group. So they are enrolling and observing and wondering. But they are not engaging like the younger ones do. I see that in my peer group and thats what I feel.

So what does that mean. The 25+ browse more but participating less while 16-25 engage as a predominant social communication and entertainment.

So what does that say for 3rdparty applications. They are still stupid because active users are 16-25. They will mature if the 25+ generation starts adopting social networking platforms for their social interaction needs. I bet that 25+ generation will do that, they are getting warmed up.. kind of learning the new ways.. isn't it always like that as we grow older.

And as more generations participate, the market matures, market fundamentals and technicals become applicable and relevant. Thats when the long tail will prosper.

my 2 c

sumit gupta

  Don Philabaum [10.14.07 12:24 AM]

Tim, thanks for sharing the research. Geller, you make a good point about the low cost of entry for anyone to create a Facebook application. I always look at the PR value this brings to small business owners. As a small business owner, I can use the Facebook branding to my customers or prospective customer to show my business is with it. As more businesses understand this, they will be able to engage and touch their customers through Facebook's platform. Low cost, no cost and a marketing communication channel. What could be better than that?

  The Pageman [10.14.07 11:56 AM]

I think it's still the Long Tail with a vengeance. Tim is talking about 5,000++ apps vs. the normal 1,000,000 items in a long tail. so if you get an app that has like 200 users, if you adjust this for scale, they should have 200x200=4,000 users daily. Not bad if you're using a Long Tail Strategy. I'm assuming here that Facebook will grow to 1M apps in the next 3-4 years as making apps becomes easier and cheaper.

  Mike Steinharter [10.14.07 01:04 PM]

Interesting stuff. What would happen if we started to imagine this in a business apps environment? What if there were an exchange/platform that enabled business apps the way Facebook enables light consumer mashups....

  Search The Tail [10.14.07 01:33 PM]

Distribution of Facebook Application Market seems to be similar to the Global Search Market Share that follows a Zipfian distribution with a thin tail (a slope of -2 and an intercept of over 60%) as illustrated in this post.
Zipf's Law for the Global Search Market Share & Rankings.

To get to 80/20 rule, the slope and intercept should be much smaller. The intercept should be about 14% for a slope of -1 as an example.

  matilda [11.17.07 08:20 PM]

Take YouTube Videos with you today announces YouTube Robot 2.0, a tool that enables you to download video from onto your PC, convert it to various formats to watch it when you are on the road on mobile devices like mobile phone, iPod, iPhone, Pocket PC, PSP, or Zune.
Product page:
Direct download link:
Company web-site:

  Phil [11.22.07 12:28 PM]

I think the reason the long tail doesn't work here is a distribution issue. In a true long tail, users are browsing with intent. On Facebook, they just browse. In a passive search, volume wins over specificity.

  Devabhaktuni Srikrishna [12.05.07 06:19 PM]

Hi Tim,

In October was it too early to draw a conclusion about the long tail on Facebook apps? By then, you may not have had a chance to look out far out into the tail either. On my blog I have a plot of the facebook apps rank ordered on the x-axis by number of daily active users as of ~5pm, December 5, 2007 - arbitrarily cutoff at apps that have 1000 or more daily active users. See

This shows there are 10,187 apps total of which about 1000 of them have more than 1000 daily active users. note this is a log plot on the y-axis to illustrate the point. Long tail: 40 million users use the top 1000 apps, and the top app is at 4 million active users. the apps above 1 million daily active users are superwall, funwall, top friends, video, superpoke, and X me. the judging from the log-scale on the y-axis, the usage appears to have a power law distribution in the range 200 to 1000. much steeper decline before the top 200 apps.

Here is my blog post



  Tim O'Reilly [12.06.07 09:02 AM]

Sri -

Analyzing only the top 1000 applications tells you nothing about the long tail.

You're basically saying that if you cut out the pesky long tail, there's a power law relationship for the top 10% of apps that does show a power law among those apps. So what? The whole point of a long tail is that the aggregate of lower ranking apps creates share worth paying attention to and that is not the case -- unless you focus on a small percent of the top facebook apps, which by definition is not the long tail.

I know this analogy doesn't exactly hold, but it's like trying to say something about the height of the populations by only looking at those who are taller than 6 feet.

When we were looking at the loglog graph a while ago, we saw the characteristic power law shape down to about the high 800's out of the 6600 apps in late October, then the cliff that showed little usage volume in the lower ranked apps - not that different what you are suggesting as of the Dec 5th data. So the point about us looking at early data causing the problem is not true. Active users remain concentrated in the top apps. And, as we mentioned in the follow-up analysis, network and viral effects seem to help make for a few big winners rather than a true long tail marketplace.

  Devabhaktuni Srikrishna [12.06.07 02:38 PM]

Hi Tim,

You have a fair point that we must verify the entire tail not just up to apps with 1000 users as in my earlier graph, and so I did just that. It shows the tail has expanded compared to October. I went back and looked at daily usage for all facebook apps as of December 6, 2007 -- total of 10205 apps listed, of which 9167 are reported to have more than one daily active user.

In a matter of two months, the shape of the log-log plot for today's data appears to have a flatter tail than than your plot of October, 2007 posted here on Chris Anderson's blog.

One interpretation is that the tail is expanding outward with time as more and more apps get written and people learn to tap into specialized interests. Approximately 40 million users use these apps — some maybe using more than one app. The he top 21 apps account for 20 million active users — only half of the total usage. The top 87 apps for 30 million, and so the remaining apps account for 10 million more.
So it's that extra 10 or 20 million users is what is worth paying attention to.

Here is updated post .


  Till [12.14.07 01:38 PM]


in a nutshell, is there any update - data, graph and/or opinion? I am curious! I know I could buy the report but I was hoping you respond here.


Post A Comment:

 (please be patient, comments may take awhile to post)

Type the characters you see in the picture above.