|More pages indexed on Tuesdays than Saturdays?|
I have a site that shows a saw-tooth graph in which weekly traffic peaks every Tuesday and is lowest every Saturday, the graph shows a steady increase in traffic over time but the same pattern holds steady for at least 8 months.
I figured it was cyclical interest in my topic but now i'm not so sure.
I've been keeping track of what's indexed in Google and I'm finding that the site has the highest number of pages indexed on Tuesdays and the lowest number on Saturdays.
It looks like the sites rankings, both in position and in quantity, drop every Saturday and raise again quickly on Tuesday... and i'm at a loss for why.
I'm not looking at a small site or a limited amount of data, the drop is significant at roughly 18-19% every Saturday. Any ideas as to what could be causing this?
Ranking changes based on day of the week definitely occur - they even get mentioned in some of Google's patents. But this is the first I've heard about changes in the total urls indexed by day of the week. I assume you mean acccording to the site: operator.
You may have uncovered previously undocumented Google behavior here. I never thought to track site: variations that closely.
Yes, i'm referring to the total number of pages returned by the site:example.com operator. I realize it's not the most accurate of tools but when monitored over a long period of time, and when compared to the ripsaw traffic graphs that repeat in the same manner over the same period, there is definitely a pattern.
Google is showing more of my pages on Tuesdays than Saturdays for this site and has apparently been doing so for 8+ months.
note: I've only tracked the number of total pages returned at noon PST daily for 5 weeks now but the pattern is predictable thus far and the long term traffic data displays the same pattern.
edit: there are a large number of factors that would suggest it is my posting behavior, choice of topic etc could be causing the swing but i'd expect the pattern to change over time, it hasn't yet on this site.
Saturdays - 3150 pages on average, it raises a little each week in line with the entire site.
Tuesdays - 3727 pages on average, again raising a little each week along with the site as a whole.
Sunday is a little higher than Saturday before the steep tuesday jump and the drop from Tuesday to Saturday is gradual, falling a little each day. I've looked for a cause and find possibilities but none that would hold over time.
[edited by: JS_Harris at 11:34 pm (utc) on May 23, 2009]
The site: tool is so inaccurate it's not funny. Just to illustrate the fact we have a site that shows 1123 pages indexed on the home page , with 723 from it's single link first sub page [ one level down ] . The only difference in pages should be one. This is replicated across all sites that we monitor.
So I couldn't rely on that at all.
As for traffic, across all sites that i have seen Tuesday is a peak day and Saturday is the lowest. It seems common across all geographic areas.
Perhaps the best observation is in the log activity. Consistantly I have noticed 2 days for site movements [ as in what happens in Google's index ]. A withdrawal [ or increase ] of content appears to commence on a Thurs thru Sunday with it's effects visible over nearly 7 days. Links appear to kick in on a Tues and do the same.
So i think user activity is different from the indexing activity that you are reporting.
Maybe others see things differently , but that's what I've seen.
Update: I compared a snapshot of my indexed pages Saturday with one taken today, the pattern held again this week, and one of the things I'm finding is interesting.
Older content isn't affected at the same rate. In fact articles over 6 months old remained indexed on both days. The newer content written over the past several months is what is moving in and out of the serps.
I assumed that content, once indexed, remained indexed but that's not the case, at least with this site. I'd say, from the articles that are moving in and out of the serps, 33% of them are between 3-6 months old and 66% of them are less than 3 months old. The site is several years old.
Perhaps fresh content is replaced with proven content under some circumstances, such as day of week etc.
Someone wrote another thread just now on something very likely related that suggests Google uses many different data points including weather, time of day and others as a metric for google adwords, I bet it applies to rankings too - [webmasterworld.com...]
edit - link to other webmasterworld article added.
[edited by: JS_Harris at 3:54 am (utc) on May 26, 2009]
Final update, just when I thought I had sufficient data to start tracking related considerations for effect Google changed gears on me, Today's traffic should have been trending downwards until Friday (though up from last Wednesday) but instead the site enjoyed a record up day... breaking a long running trend.
I suppose nothing ever stays still long enough to fully evaluate online, ah well.
Have you put your terms into Google Insights, limiting to the last 30 days?
I'd be curious how that works out as per day of the week. If you type in "sex", for example, you'll get a noticeable spike every Saturday (less so for "love" BTW but still noticeable). Searches of "jobs" peak on Mondays, sometimes Tuesdays.
So who cares? You said it's not a function of search volume in your opinion.
I was wondering a couple of things
- if you do your site: when global search volume is high, would it return fewer pages just because it won't spend as much time looking in the index for each request. And vice-versa when volume is low.
- if you do your site: search when keyword specific traffic is high, will it tend to spend a little more time on that query on its big days?
So if you repeat your query for some huge, high-volume jobs sites and some huge, high volume sex sites (it's just research honey, really, our income depends on this), would you see anything that you could match up based on Insights.
A fun update - I just noticed that on the analytics login page a sample graph is displayed with the text on the left... It shows the exact same behavior I described above repeated for the sample month. A sharp rise followed by a slow decline each day of the week and it bottoms out sharply before bouncing to a new high, rinse and repeat, the graph on the login page shows 1 month and 4 repetitions of the same behavior.
Edit: yes, I ran the terms through G insights and the pattern doesn't hold at all.
[edited by: JS_Harris at 9:31 am (utc) on June 8, 2009]