I don't think there's a short-cut to differentiating yourself there: you just have to run experiments and not use that language to roll out new features (ie, ne

I don't think there's a short-cut to differentiating yourself there: you just have to run experiments and not use that language to roll out new features (ie, never make an experiment permanent at the end, but roll it back for a few days/week/month, discuss it, and then redeploy). I would actually trust a website that made too big a deal about how they weren't like those other guys less, because that's an old marketing trick. Same way you build a reputation for anything.