7 ms·
A/B test them. Pick something arbitrary that looks pretty good for the initial version (like the initial top-ads color for Google's May 2010 websearch redesign
by nostrademons 15y ago
A/B test them. Pick something arbitrary that looks pretty good for the initial version (like the initial top-ads color for Google's May 2010 websearch redesign was "Cosmic Latte", the average color of the universe). Make it configurable in the source code, either through template variables or frameworks like Sass. Then vary one parameter at a time and measure your business metrics, and do what the data tells you to do.
It's important for designers to get the big-picture stuff right, like page layout and interaction model. For pixel-perfectness, just do what the data tells you.
- kurtvarner 15y agoI agree about the importance of A/B testing, but it's not realistic to split test details like 16pt vs 18pt font, slight gradient changes, small text positioning, etc. But you're right about designers understanding and focusing on the big picture. It's just hard for my mind not to get lost in the details.
- nostrademons 15y agoWhy not? Google does that all the time. Those are perhaps the easiest details to test. If you're using a framework like A/Bingo, it's literally one line of code.