trendpreneur

Sometimes the behaviour of the masses is all you need to get some fairly solid recommendations. The example I’ll use is World of Warcraft addons, but it really holds true for any sort of platform where there are a vast number of third-party modifications — you could use the Xbox Live Arcade, even, or web widgets. However, addons are an interesting case for a couple of reasons.

Firstly, the addon space is pretty crowded. An addon is a Lua-scripted modification that’s written by a third-party and enhances one or more aspects of the game. Thus plenty of enterprising hackers have put together things they specifically wanted to code, with the upshot that there are several options to choose from in pretty much every category. “What’s the best addon for…” is a common question, and in certain categories there isn’t a clear winner.

Another side-effect of this is a search problem — even though some addon hosting sites use tags and categories, it can be really hard to find something that does just what you want without human intervention. “I want an addon like X, but that does Y”… sadly, that’s not a language a category list can understand.

Secondly, as the game updates, addons are continually updated, so users tend to download their favourites multiple times. Of course, some are more diligent than others, and some people use deprecated versions of addons because they like them (until they break, that is). However, this measure of downloads is a really easy way to see where the user favourites lie.

Crowdsourcing the solution

To my mind, there’s an obvious way to combine these two properties: an addon hosting service that also taps into the user’s download behaviour — ideally, you’re the only source of downloads for the user, for more accurate behaviour tracking. Providing a trusted standalone program (even better, one that reads to see which of the installed addons the user actually *uses*) is an easy way to make the user experience better at the same time as improving accuracy. Updating all of a user’s addons can take some time, also, so you can add in something that encourages the user to give addons star or thumbs ratings while they wait.

By tracking which addons people actually use, and tracking how often they update/download them, a picture of the overall behaviour and preferences builds up. Add to that the usual social features such as tagging, reviews, comments and a recommendations engine — “people with Grid installed usually have oRA2 as well” — “people who browse Bob’s UI pack tend to also look at Joe’s and Sam’s UI packs, and eventually download Joe’s”. Then put in lots of user-friendly features such as favourite lists/update alerts (which you can also use to track preferences!), top-10s and a “help me find” wizard, human-powered thread or even volunteer help base, an IRC channel (also tracked for addon mentions, of course)…

Eventually you’ve got this fantastically detailed picture of what people actually use, and you can throw that right back at the users in the form of popularity and opinion filters that really show “this is the best addon for x (but you might like this one too)”. It’s also going to be easy to see which addons 90% or more of the player base use, something the game manufacturer would surely want to know (ok, they can probably get at that information themselves, but not with the power of a community behind it).

Maybe the Big Brother approach here would put people off, if they realised that their behaviour was being observed in this way. It’s becoming increasingly common to watch what users do, though, and in this case it clearly results in a superior product to the user — a much easier way to navigate the mess that is third-party addons. The same principles can be applied elsewhere to much the same end, as well, and in many cases tracking simple things like the number of downloads (to provide a ‘most popular’ list) is already happening.