What’s Causing Google To Shuffle?

28 Sep 2010, Posted by admin in Google, 12 Comments

What’s Causing Google To Shuffle?

For the last few weeks Google, in the UK at least, has appeared to have been shuffling the listings around for very high volume, very generic terms. Queries like ‘car insurance’ have seen massive fluctuations, often several times a day.

It’s very difficult to say with any certainty what’s happening here, but the big shuffles remind me a lot of the Vince update from last year – the theory I had for that was that (amongst other things) Google was shuffling the listings around to check how well each result performs when put in a different position. Does Site A get more clicks than Site B when in position #1? Or #5? Does Site B have fewer people returning to the SERP shortly after clicking on it? (Visitors returning to the SERP soon after clicking the link could be a sign that the result didn’t help them for that query).

Either way, it looks like the shuffling we saw after Vince has just become more aggressive, with wilder and more frequent fluctuations.

While it’s never a normal SERP at the best of times, the results for ‘car insurance’ are a good example of the shuffling results, with changes happening several times a day. I often find the lower positions of page 1 to be quite interesting, as I have a theory that Google often tries to keep the top positions a bit more stable so that the user experience isn’t too disjointed.

I asked on Twitter recently to find what search listings people were seeing in positions 8, 9 & 10 of for ‘car insurance’ and found that there were at least 3 different SERPs that people were seeing – just for those bottom 3 positions. Asking Twitter the same question the next day showed, again, 3 variants of the SERP – all different from the day before. (Everyone that helped me out on Twitter – thanks very much).

This implies that Google, if they’re testing things like click-through rates and satisfaction rates, are likely to be testing several variants of SERPs across different data centres, presumably to help shorten the time it takes for them to find the perfect 10 results to answer the user’s query.

I’m still in the process of testing this in a bit more detail, but it’s worth keeping an eye on. If you’ve noticed any SERPs that have started to shuffle massively (I’m adopting the name serpquake from now on), particularly in the last few weeks or so, it’d be great if you could leave a comment.

Flickr image from Ralphunden.

Promote Post

Enjoyed this post?


September 29, 2010 1:56 am


Interesting stuff, I’ve had similar thoughts. I’ve got 4 websites all different verticals that for the last 4 weeks in a row dropped out of google on a Tuesday and then returned on a Friday, 4 TIMES!!!!! the only things the same are similar linkbuilding techniques.

September 29, 2010 8:40 am


Fascinating post and I like the name serpquake. I have been experiencing the exact same thing with a number of my most competitive key phrases. I will do the search in three different browsers and get three completely different results. If your theory is correct it will be interesting to see how the listings change after the serpquake.

September 29, 2010 8:59 am


We’ve been seeing exactly the same thing for high volume, highly competitive key phrases.

I have had three different result sets by using 3 different browsers at the same time. It looks like G is using cookies to keep the results stable for each user for a day or so. If you use different browsers you can see a variety of result sets simultaneously.

September 29, 2010 9:32 am


Interestingly, page three doesn’t seem to move whatsoever. If they were algorithm changes, one would expect even minor changes to page three.

September 29, 2010 10:00 am


Bowdeni – that’s a really interesting point, I’m starting to track some SERP changes so I’ll look into that more.

Brad & Sam – Really good spots, the cookie idea makes a lot of sense.

September 29, 2010 7:07 pm

Lloyd Cooke

Interesting point about the data centres.

I remember in the insurance vertical about 12 months ago there was a very similar test and looked to be CTR led as Google was trialling different types of companies in the #5 position for home insurance and car insurance.

For home insurance the following types of brand/company offering home insurance was ranking at #5:
– Over 50s insurance company
– Supermarket brand offering insurance
– aggregator (large brand)
– aggregator (small brand)
– standard well known brand insurer

For motor insurance the following types of brand/company offering home insurance was ranking at #5:
– breakdown cover brand offering insurance
– over 50s insurance company
– supermarket brand offering insurance
– standard well known brand insurer

Would be interesting if they also took the bounce rate data into consideration on this to0?

September 29, 2010 9:00 pm


Thanks for the comment Lloyd – I completely agree with the click-through rate idea, although it’s not really possible to prove it, it does make the most sense. It depends on how you mean bounce rate though – I assume you mean people bouncing back to the search result (which would suggest that the site didn’t answer their question), which I also totally agree with – although again, can’t really prove. This post from Blogstorm shows a Google engineer talking about it and suggesting that that’s the main change they made with Vince.

Not to be confused with just any bounce rate though (like if they immediately went off to Facebook or anywhere other than straight back to Google) – which Google wouldn’t be able to see without looking at Google Analytics data, and I’m almost certain they wouldn’t do that.

September 30, 2010 8:41 am

Phil MacKechnie

We studied this with some considerable data from car insurance SERPS 6-12 months ago for all of the top 50 results. The data indicated clear ranking bands for top generic terms which were either very stable or susceptable to great flux.
Positions #1-3 – solid rankings, very rarely altered for months at a time
Positions #4-6 – rankings altered only occasionally, from time to time replaced by pos# 8-10 for days at a time which suggested testing of the factors discussed by @lloydcooke
Positions #7-10 – these rankings altered the most, both up the page and off the page to pos #11, #12.

The data was pretty conclusive at the time and really helped set clients expectations with DATA, removing opinion and theories from discussions. Its probably time to run the same tests again across this SERPs to see how it looks again. Wouldbe wise to pick another couple of examples. ‘car insurance’ as many people know is a particularly special SERP in the UK which gets a lot of attention by SEOs and Google engineers alone. Testing on regular SERPs often doesn’t pick up some of the latest algorithm tests.

September 30, 2010 10:15 am


Thanks for the comment Phil – I completely agree, “car insurance” on it’s own has always been quite an unusual SERP compared to others. I’m currently running ranking reports for a number of different search terms, of varying levels of competitiveness, so in a few weeks I should hopefully have something a bit more concrete to share, although I think some of the more likely theories (click-through rates, satisfaction rates) can only be theorised about rather than explicitly proved.

September 30, 2010 11:34 am

Peter Young

Interesting insights from Phil and Lloyd.

With regards to the points about CTR, I remember Google saying they used the search behaviour data for ranking Youtube videos during one of the sessions down in London about a year back.

Continuing that line of thought I would suggest the likelihood of similar technologies porting between Google products is highly likely, and usage in the mainstream serps would only make sense based on the behaviour outlined above (and from an IR perspective), and the fact Google already include personalised data within serps + filtration if there is any merit to Malcolm Coles post this morning.

September 30, 2010 11:49 am


I’ve seen massive drops and position changes over the last few weeks. With no real consistency between the sites being moved around other than them being new sites. I guess they are shuffling things around to see where they are best suited in the rankings

October 16, 2010 3:43 pm


Yes, this is indeed very interesting. I think that Bing is making a big impact on what Google are doing at the moment. One thing I am trying to get my head around is that my site has quality links to it and when i check it I find that Google cannot find one link? However, the other engines are. Movement is up and down for a lot of SERPS and this must be causing a lot of SEO consultants a lot of concern. Also, why do we need instants and how do you know what settings to have on Google settings (moderate,safe etc.) because you get diferring results when you change your parameters. Great post.

Posting your comment...