If you pay attention to the icons of games in the top grossing charts on the AppStore, you will have noticed the Shouty Men in Hats phenomenon.
For a while, all the top grossing strategy games had for their icon a man. Wearing a hat. In an open mouthed shout.
Yesterday, that all changed. According to Twitter, Supercell closed the mouths of their shouty men on both Clash of Clans and Boom Beach (but not, by the look of things, Clash Royale).
The identikit icons are a classic example of a local maximum, and one of the perils of data.
If I want to get to the top of Everest, the tallest mountain in the world, I can follow the simple instruction “Always go up”. I will soon find myself at the top of the nearest molehill.
I have found the local maximum by following a simple dataset. But to actually get to the top of Mount Everest, I have to go down and back up again, over and over again. It requires belief not in the data, but in a combination of data and a wider-understanding.
I use this analogy in explaining why data-driven design is weak, but data-informed design is amazingly powerful. In short, we still need designers and creators to make new leaps of faith, to come down of their molehill to start climbing the foothills, and eventually mountains.
The icon designers at Supercell and Machine Zone have tested their icons. They have found that “shouty men with hats” get the best conversion rates. So they use those icons, and everyone else copies them, and soon all icons in the Appstore look the same.
One day, someone will come along with a radical new design (no, not just closing the mouth), and they will get oodles of free downloads. That will be refined, and copied, until again it is local maximum. And the cycle repeats.
Local maxima like this show the dangers of relying too much on data. It is a powerful tool for optimisation. But it can also reach the point where it stifles growth.