Time of day rankings changes

[Sgt_Kickaxe]

I'm seeing some time of day related rankings changes this week. Today for example from midnight to noon traffic was roughly 40% lower than the average and at exactly noon it returned to within 1% of average.

Yesterday it seemed to work in 4 hour bursts. Between 30% and 50% less traffic 4 hours at a time followed by near normal four hours at a time.

I can't attribute this to anything I've done, the site is 6 years old and unchanged in quite a long time. Traffic sample size is typicaly 14,000 visitors a day with 31,000 pageviews so I don't see it being a small sampling issue either.

This is happening with Google only.

[tedster]

I call it "day parting" and it's definitely on the rise. I first saw it over two years ago, but rarely. Now it seems to be quite common, compared to back then.

There are also day of the week ranking changes, and time of the year changes.

Time of day ranking changes seems to be a mechanism that "traffic throttling" uses. So if you are diagnosing a traffic question, definitely DON'T check the ranking at the same time every day.

[Robert Charlton]

There are query variations (and thus traffic variations) that definitely relate to searcher behavior at different times of the day or days of the week. These are for the most part easily explained... often correlating with business hours, eg.

Why Google should also vary rankings at particular times of the day, though, is something I've heard reported and I accept, but I still can only guess at the reasons. Seasonal and time-of-year ranking changes are perhaps more clear in their intent.

I've seen what are obviously test serps occasionally rolled out on weekends, but beyond that haven't understood why some reported patterns happen as they do.

On something as regular as 4-hr cycles, it might be that Google is trying to normalize new sets of data inserted into serps, much as they would try to normalize Universal results. It might even be that Google is trying to find out why traffic is time-dependent by putting different types of sites in different positions, in order to look for different sets of correlations.