by: danah boyd
The number one justification i get for Internet Explorer-only support is that 90% of the population uses it.
Let's assume that to be true (even though only 52% of this blog's readers use IE5 or 6). This argument rests on two assumptions:
1) An individual uses IE (and ONLY IE) on all computers that they use.
2) The only browser that matters is the individual's browser.
When users cannot use an application as they move between work and home computers, between personal and school computers, etc., they get disincentivized. Yet, that's a minor problem compared to #2. When it comes to social software, i'm not just concerned with what browser i use, but with what browser my friends use. I may not be concerned directly, but i need them to play along too to get validated and to make it fun. I don't want to invest time and energy into making profiles or blogs that my friends can't access for functional reasons, especially if there are alternatives that everyone can access.
You need cluster effects for social software to work. I need to be able to convince my most exploratory friends to try it with me and i need them to get super excited about it. Once i get them going, then i can convince the rest of my friends to follow along. If i can't convince them, then i quickly lose interest and stop trying to convince everyone else in my social world. Not only does this make it hard for me to play along, it makes it hard for my close friends that i turned on to play along. Because if i lose interest, why should they keep spreading it to their friends? Etc.
For entertainment, let's play a probabilities games… Let's assume an even distribution of IE use (which is not true) and random friend connections. Let's assume the average teen has 40 AIM buddies (low), but that only 10 really matter. In other words, 10 specific people are a critical baseline for my desire to become an active participant. (Note: the self-motivation to try it about early adopters does not take into consideration whether or not my friends will play along.) There's a 34.9% (.9^10) probability that all of my close crucial friends are on IE. Let's say that i'm in that important 35%. For it to take hold, all of my friends need to participate and pass on the enthusiasm virally. The probability that all of my important 10 friends are also in that critical 35% is… TERRIBLE (assuming random friendship connections). As network effects take hold and interest spirals, there will be critical nodes who simply don't participate for structural reasons. That is bad bad bad for significant growth and sustainability.
Of course, in reality, browser use is not evenly distributed, friendship networks are not random and it's not clear exactly how many crucial people one needs to participate. (Translation: the probability game was for kicks – a real analysis would require modeling network spreads and calculating stickiness.) There are likely to be quite a few IE-only clusters, but there are also likely to be quite a few clusters where crucial nodes use Firefox/Safari. (There are also likely to be a few where there are other browsers, but frankly, these are typically the geek networks that most mainstream developers are happy to write off.)
The important thing is that when you think about browser-access, you cannot simply think in terms of "90% market" because there's a decent probability that many of those 90% have critical connections to people who are in the 10%. You need to think in terms of clusters, not individuals, because it is clusters that will make your application work. People participate when all of their friends can.
Corporations force this through regulation software, but this is not how consumer markets work. Launching a beta of AIM Pages on IE-only is foolish at best. Sure, a lot of people will try it, but if their friends can't play, they won't really get into it. Meaningful activity won't spread unless entire clusters can play along. (Trying it out by creating an account is not the same as being active.)
Getting social applications going requires a baseline…. That baseline is that everyone can play along so that there's no structural barrier to network spread. This is why mobile shit is so hard to get off the ground. This is why getting people to download applications for social interaction is such a barrier to participation. Replicating this problem on the Internet is foolish at best. It doesn't matter if you're launching in beta – first impressions really do matter. If you're targeting an audience that's IE-only (like corporations), go for it. But if you're trying to go after a mainstream, younger audience, you're being idiotic if you think you can get away with not supporting Firefox or Safari. (And besides, if you're AOL, what on earth are you doing supporting Microsoft hegemony?)
Original Post: http://www.zephoria.org/thoughts/archives/2006/05/13/cluster_effects.html