I compared several at once in my own time, instead of going on impressions. What I found is that the ratio of active to dormant profiles accounts for most of the variance I saw. The difference was visible reliably even when I varied the time of day. It cost me nothing to verify ultimately, whether others found the same thing.