Personals hourly activity

Following my last post curiosity got the best of me so I had to investigate to see what the hourly usage was like on my sites.

Looking at unique page views it’s very clear that everything is very much how I would expect it to be. Hourly usage is at a peak from 5pm up to midnight. It dips while everyone sleeps and then picks up throughout the day.

Though I’ve not plotted payments over this, the trends do very much follow what I have observed. There are payments very soon after midnight, then little during the morning where it slowly picks up and hits a peak post 5pm which continues through the evening.

At the moment new traffic comes evenly from organic and paid campaigns, with no throttling our end, the distribution across the day is purely from when people choose to access the site or click on a link/ad somewhere.

I aim to have 30% new traffic to the site on any day. Unless we targeted very aggressively traffic at off peak times and stopped completely during peak different times it would not affect the graph much. And we’d have to ask ourselves why we would do that - is the ROI on members in the middle of the night better than those that will come to the site during normal hours? Again, it’s not tested but I seriously doubt it, so much so I won’t bother investigating just yet!!

I’ve not looked at day of the week activity, partly because it is far to easy to influence with newsletters which go out. If you send out a newsletter Wednesday you’ll be biggest spike of activity from this newsletter soon afterwards. This is particularly true on dating and community sites. What would be more interesting to know is on which day the same mailing gets the highest response….

If you do start looking at information like this to optimise advertising or any other call-to-action initiatives it’s important to realise that the moment you start reacting to trends, such as hourly activity, days members will read emails, you will also skew your results, so you need to segment and so that you can either test alternatives, or have a “normal” data set to compare to.