Wednesday, 16 July 2008

Some Things You Should Know Before Developing a Social Site

Designing and building a social site, or “online community,” can be painful. When the client doesn’t understand what leads to success in this space, which is almost always the case, it can be excruciating. Marketers breeze in with MySpace and Facebook sized expectations fueled by the misconception that simply throwing up a mess of loosely themed, content-free social functionality (surrounded by a high wall) will somehow drive “connection” hungry users to their door. Not surprisingly, more often than not these projects fail. And, of course, the designer gets the blame. If the designer’s feature set or IA or visual design had been slightly better, the community would surely have thrived.

This post is a bit of tough love for those embarking on social site development. What I’m about to suggest might sound like the rationalizations of a burned designer, but understanding it can really help manage client expectations. The online social space just doesn’t work the way most of us think it does. Beyond a certain point, things that designers have direct control over tend to be minimally determinative of site success. That is, once your site looks credible/"genre-appropriate", has a non-scary modicum of differentiation and honors the default category expectations, the designer’s direct impact on success dwindles rapidly. Why? Because success in this space is largely (but not solely) driven by subtle, nonlinear social network-based phenomena that have very little to do with the qualities of the actual site itself. Good design is clearly necessary for marketing and sensible development, but it's nowhere near sufficient for success.

Three closely related phenomena seem particularly important to social site success. The first two are what’s called “externalities,” which for our purposes boil down to a person’s behavior being partially determined or influenced by other people’s observable behavior. Bearing that in mind, here are the three tricky phenomena.

1. Informational Cascades

Cascades (for short) are what’s called an informational externality and they impact the decision to join a community. The idea is that it’s cheaper in time and attention (and thus still strictly “rational”) to join a community others join than it is to waste energy mulling over the pros and cons, regardless of your own judgment of community value. Learn from others and, if they’ve done okay, you’ll probably do okay, whatever your personal assessment of site worth may be.

So after a community platform gets some traffic, most users shut off judging for themselves and join because certain relevant others have. Differentiation or design based incentives to lure folks in only really matter for the first n users, where n is relatively small and extremely fickle (so-called early adopters?). After that, the join decision is often based on 1) the fact that others are doing so and 2) the assumption that they must have good informational reasons for doing so. Thus 3) new users often disregard private information, meaning that 4) the information of all but the first n becomes effectively irrelevant. VoilĂ , exponentially diminishing returns on design outlay (and testing!) beyond a relatively low threshold.

2. Network Effects

Like cascades, these are also a type of “externality,” a consumption or production externality. Network effects impact the perceived value (or utility) of the site both before and after the join. In a nutshell, the value of social functionality increases nonlinearly (some say logarithmically, others exponentially... the jury’s still out) with the total number of users.

Actual details from economists and “theorists” in discussions of network effects are mostly fudged idealizations, but the kernel of truth is that people consider social functionality valuable only if some significant proportion of people (usually friends and acquaintances) are using it. So, your online community might have the coolest social functionality, but if you never managed to get a significant number of folks to make that initial commitment of time and attention (say by having decent content or non-social functionality to get traffic in the first place), you’ll never get off the ground. And without people, a community ain’t... and never will be.

3. Path Dependency

This one tends to be the result of the other two processes working together. Often, the situation the world ends up in is the result of a large number of historical moments in which agents choose between a couple of alternatives. When we’re dealing with externalities like the one’s mentioned above, these small choices in the past can have a startlingly powerful impact on the present. A notorious feature of these path dependent processes is that small (really small) changes in the decisions made by individual users in the past can have radical effects on how the present turns out. Winners could have been losers but for some schmo joining one site and not the other. From this decision, cascades can occur, with incipient network effects following shortly behind. Presto, a path dependent process with significant potential for sub-optimal lock-in.

Some Elaborations and Lessons

Big words, scary ideas, but what does it mean? Below are some simple, realistic consequences of the above ideas with some suggestions for actual development. Keep these in mind when embarking on a social site project and let them guide your expectations and success parameters.

1. This one’s been said a lot, but has never really sunk in. Provide some reason for people to use your site even before there’s any socializing. You need content or independently useful, non-social functionality to get to the point where the externalities discussed can take hold (unless your product is just a cool bit of standalone social functionality like a “widget”).

2. Beyond a certain default determined threshold, fanciness of design and coolness of functionality are only really important in terms of innovation and differentiation not conversion. In other words, they won’t guarantee success (in traffic or use) no matter how relevant they are. They can get you noticed, which may start a cascade among those with a low switching threshold (early adopters?), but if the network effect threshold isn’t reached, you’ll remain niche (sorry, Virb). Remember, though, that trafficking in early adopters is particularly dicey given that they've especially low switching thresholds... they'll abandon you just as easily once something newer and cooler comes along. If they abandon before the cascade can reach a scale large enough to overcome the higher non-early adopter switching thresholds you won't reach your tipping point.

[ASIDE: The iPhone isn’t necessarily a counter-example. It does in a new space (mobile computing) what Apple does everywhere (simplify and amaze) and relies on Apple’s existing converts and proselytizers to fuel uptake (create a cascade). It’s success is the result of an already converted base expecting and getting really cool design, not of really cool design converting all comers regardless of prior commitments.]

3. A hallmark of path dependent processes is that they tend to be “irreversible” to a certain extent. For our purposes, this means that once your users have made an investment of time and attention, switching isn’t as easy (costless) as marketers think; switching isn’t zero-cost and this can create a kind of inertia in users. In other words, just because you build it, that doesn’t mean they’ll come. Users must perceive value significantly greater than the switching cost of time and attention. Without that initial threshold of users (network effects) or initial visible rush (potential cascade), there’s little perceptible value and no informational precedent that tells them joining might be a good idea.

4. The value of testing diminishes rapidly as uptake increases; it’s value is effectively nil beyond the first n users, where n is the “tipping point” (the point where a cascade occurs and network effects can start to be felt). Also, it won’t guarantee success, even if the product tests well. All testing really does for you is ensure you have made something that honors current design and functionality conventions, something that’s not scarily new. It will never guarantee success because you still have to provide extra value to overcome switching costs, no matter how low. And testing certainly won’t allow you to create new competitive landscapes through innovation. So, definitely test but with diminishing value and limited ends in mind. (Remember, this is only really applicable to social sites... transactional sites, etc. tend to get more value out of testing)

5. If the success or failure of online communities is a path dependent process, then copying the design of other social sites in hopes of copying their success makes very little sense. Of course you need to honor conventions, understand defaults and identify areas for relevant differentiation in the space, but beyond this you’re not really generating value (and it’s really lame to boot). Qualities matter insofar as truly bad or scarily unconventional sites won’t be in the running. However, the predictors of success typically aren’t site qualities (the various competitors all do pretty much the same, category-appropriate things), so imitating qualities of successful sites is no guarantee, only a good bet. Finally reproducing the winner is silly given non-negligible switching costs. People won't jump unless your offer exceeds the switching cost set by the current site and its value. Don't offer what they already have because there's no incentive to move.

6. It’s not always the case that the best community (by whatever standard) wins. Informational externalities (cascades) coupled with production/consumption externalities (network effects) often lead to a “sub-optimal” option taking the win. Lock-in as it’s called is clearly observable in situations like the maintained dominance of Windows, the market triumph of VHS over Beta, or (everybody’s favorite) the QWERTY keyboard. Of course, your product needs to be competitive, moderately differentiated and (if this is your thing) “innovative” in a non-alienating way. But developing and (especially!) testing at great expense your new category killer won’t really put you in a better position to recoup your investment. At most it means you’re a standout among the crowd of competitors. Without significant marketing and some lucky breaks, all that extra work simply sets you up to look new, feel cool and keep users IF you get them in the first place. It’s a set of sad facts that the best rarely win, first mover disadvantage is very real online, and the just good enough often get the biggest return (I’m talkin’ about you, MySpace).

Well, those are my observations. Hopefully, you can now embark on your development with clearer vision and a better sense of what can and cannot be accomplished through design. Anyway, hope it wasn't too scary...