User Interface Metrics

Last spring, I decided that the next game I work on should have user interface metrics. By this, I mean I would gather metrics on how the user interacts with the game (where they move the mouse, where they click, when they click, etc.). Given a deterministic game engine (which is how I generally like them to be), one could theoretically playback the UI metrics log to reproduce the exact game session. While I hadn’t spent a lot of time thinking about specifically what information I would gleam from this data, it sounded like a useful (and interesting, and lightweight) metric.

A few weeks ago, I saw this cool utility of TechCrunch: Clickdensity. With just some simple JavaScript inserted into your web pages, Clickdensity builds heatmaps for your web pages. Very cool. I saw another bunch of script earlier today (can’t find it now, though), and just now saw a mentino of CrazyEgg, which does the same thing as well.

And to complete this meme, today on Gamasutra was this fine article on RTS interfaces, Too Many Clicks! Unit-Based Interfaces Considered Harmful. Around page 5 or 6 he makes the suggestion of user interface metrics, describes weighting them, and even provides some references on prior research in the topic.

How’s this more broadly useful to games, particularly the heatmap tools? Well, first of all, games are all about interface design. A game’s UI can make or break its usability, and by extension, its fun (as opposed to it becoming a chore). A heatmap can reveal those parts of the UI where the players consistently click with some expectations but are left wanting. They can also indicate organizational problems: if you have a big cold spot in the middle of your UI, then it’s a likely candidate for simplification. Or, if you find a stray UI element that is consistently as hot as your main UI elements, then perhaps the two need to be better integrated (or at least brought spatially closer).

Given the proper UI metrics, one could look at them plotted over time, specifically their frequency. The developer would want to closely investigate any spikes (or valleys) as they are potential indicators of pacing issues, or confusion on the part of the user.

[Update: I found the other clickmap/heatmap site I mentioned above: The definitive heatmap.]

  • Patrick

    That was a good article. Notice how the high-level abstractions proposed for improving RTS UI involve interaction with human-esque agents. On that note, another approach to UI improvement is blind testing, which generates similar data emphirically. Either way, UI is key, I consider it my primary speciality next to dealing with rude geniuses.

    I think Fianna has a good head start in this regard, since the interface involves clicking on an object, then holding an arrow key for a period of time. I don’t think it can get much simpler than that, one click, one button. I’ll look foward to testing a prototype with random people.

    Now go comment on my blog dammit!