7-pack vs Blended Local Results – Split Testing?

Why does Google display a 7-pack for some queries, and blended results for others?

I was testing out a search query from one of my Local Citation Finder users for ‘santa rosa automotive repair‘ and I came across an interesting case where in one search I would get 7-pack, and then I’d search again and get blended. I’ve been perplexed about why Google shows a 7-pack for some queries and blended results for other queries ever since David Mihm asked about this on the 2011 Local Search Ranking Factors Survey, so I played around for a while trying to identify what might be going on.

I was able to duplicate the results multiple times. If I searched one way, I’d get the 7-pack. Make a slight variation, and I would get blended. The difference in how I searched to get the two variations was so minor, that I was surprised to get a 7-pack on one search, and blended results on the other.

Screenshots with date and time recorded:

7-Pack (click to view details)

7-pack-thumb

Blended (click to view details)

blended-thumb

Here’s the play by play:

Search 1:

  • I put my cursor into the address/search bar in Chrome, and I type ‘ santa rosa automotive repair’ and press enter.
  • I get a 7-pack.

Search 2:

  • I put my cursor into the address/search bar in Chrome, and press enter. Exact same query.
  • I get blended.
  • Huh?

Search 3:

  • I press the back button.
  • I type the search phrase into the address/search bar in Chrome again, and press enter.
  • I get a 7-pack.

Search 4:

  • I leave the search term in the field, and I press the search button on the results page.
  • I get blended.

 

I was able to duplicate these results about 10 times (I was trying to get good screenshots).

Finally, about 4 minutes later, I was no longer able to get the 7-pack. I would get blended results no matter how I searched. So, what’s going on here? First of all, I think it’s possible that Chrome cached my first query, and that’s why searching that particular way continued to bring up the 7-pack. Searching a slightly different way somehow told the browser to run a fresh request from Google’s servers. I don’t know very much about how Chrome’s caching works, so this is just speculation.

Assuming it was just cached, why did I see a 7-pack in one instance, and then blended results a moment later? I have a couple theories, and I’d love to hear your thoughts in the comments:

  1. Split testing. I just happened to hit it at the right time when they were just switching the test over to a new version.
  2. Query volume. Perhaps low volume queries get a 7-pack, and when I was hitting that search term, I just happened to catch it right as it passed the threshold to trigger the blended results.

I’m putting my money on split testing, because I have seen some queries that would definitely have high search volume return a 7-pack. For example, check out these two searches:
http://www.google.com/search?q=seattle+lawyer (7-pack at time of writing)
http://www.google.com/search?q=seattle+lawyers (Blended at time of writing)

Why Google decides to show a 7-pack sometimes, and blended other times is definitely still a mystery, but this is a case that seems to indicate, to me at least, that it’s just split testing.

What do you think? Any other theories about what’s going on?

 

AUTHOR

Darren Shaw

Darren Shaw founded Whitespark in 2005. The company specialized in web design and development, however, Darren's passion and curiosity for all things local search led Whitespark to focus primarily on local SEO in 2010 with the launch of the Local Citation Finder, followed by the Local Rank Tracker.

Follow Me on Twitter

0 comments on “7-pack vs Blended Local Results – Split Testing?

  • When Google first rolled out the blended results and I saw the varying styles based on query I thought perhaps it was split testing. However the more I see it, and the queries I see it on I am lead to believe this is more due to query volume and search intent over anything else.

    Small volume queries and queries that deliver fewer clickthroughs to locations seem to be void of the blended results or any local results even for local queries. I would assume that as more people become accustomed to using Google Places and the blended results we might see this trend change in some of the lower volume queries, but I think it will honestly come down to search behavior for that query and not the overall behavior of users on Google.

  • I think sometimes you might get immediate personalization – if you get a 7 pack and don’t click on anything in it, the next time they serve you up blended listings to see if that’s what you want, but I do think split testing is always going on too.

  • One other option would be that Google thinks blended is be a better result for a repeat search.

    I think the whole 7-pack/blended is a great mystery… curious to hear if others have ideas.

  • Tried replicating your tests, got 7-pack each time for santa rosa automotive repair and got 7-packs for both seattle lawyer and seattle lawyers. Tried in both Chrome & Firefox

    Some split testing does make sense, but for some queries, there does seem to be some level of consistency. I think between your testing and Linda Buquet’s research, someone will one day get to the bottom of this!

  • While determining the factors that cause the change is challenging (although it seems location of the searcher is one factor…), I strongly believe this is neither testing, nor random. The changes only appear during Google Places data updates, i.e. they are caused by true data changes on Google Places itself. Therefore, one possible theory could be that results depend on the relative strength of local search factors, such as citations (and especially the number of citations under a specified category) vs. companies’ websites. The number of listings associated with a specific category compared to the total number of listings for relevant categories might also be a factor.

  • I agree with Mike Wilton on this one… i believe it’s “due to query volume and search intent” on which determines whether you get served up the ‘pure’ results or the ‘organic’ results.

    I also believe they have kept the two for some very important reasons.

    Google always wants serve of the best results for searchers. That is ALL they care about. Every change they make is to improve the searching experience for regular folk – not us SEO’ers 😉

    The reason I believe they created Google Places and the 7 pack to begin with was they knew a lot of searches had local intent and they knew a lot of local businesses did not have websites, nor were representing their local businesses well online.

    So Google took it upon itself to create GP, so that they could serve up the proper results searchers were looking for.

    Now flash forward years later, after tons of GP spam and working with a GP algo that doesn’t really work… they decided to move back to an algo that does work, the blended search results (good regular SEO – yeah!) …. which works for most lcoal search results for searches that are for businesses inside larger cities. (because the query volume is high enough that a lot of small businesses have begun to market their local businesses online).

    Even though the new blended results works better for both searchers and Google, they couldn’t abandon the old 7pack quite yet.

    The reason being is not all cities and towns, etc have a high enough search volume, thus resulting in less local businesses having a local presence online. If local businesses don’t have a presence online that is best for searchers, what is Google going to do….? That’s right, give them the GP 7-pack… show the ‘pure’ search results because it’s a better experience for searchers.

    I believe they use “query volume and search intent” as deciding factor b/c Google can only make algo changes and that is the closest way for them to serve up the best local results for searchers.

    Now I also believe there is a little ‘split testing’ and ‘personalization’ involved in the mix too, making very difficult to fully know what the hell is going on.

    That’s my ‘theory” and I’m sticking with it! 🙂

  • I like your analysis on this and I agree that query volume might be the discerning factor. I’ve also seen examples where if the sites ranking organically did not have enough anchor text optimized inbound links, the result would show a 7 pack instead of blended results. That said, I’m seeing blended results for ‘most’ of our clients.

  • I experienced the same results with one of my client’s keyword (in a competitive industry, with relatively high search volume), but the strange thing was I would repeatedly test and get a 7-pack in the morning and when I tested the keyword by end of day, I would get a blended result! It was most bizarre and it lasted for almost a week. The results for keyword have stabilized after the recent Panda update however. I agree with Mike, on the fact that it may actually have to do search behaviour of a particular query.

  • So following your tweet to me on my theory I did some more digging. Most of the work I do involves businesses that provide services (e.g. lawyers, doctors, photographers, etc.) and in my realm it seemed as though volume was the difference. The example you gave me was a restaurant. I did similar searches (e.g. restaurant, grocery store, clothing store, bar, etc.) and all of the queries seemed to feature the 7 pack.

    Could it perhaps depend on the type of business also? Perhaps if you are just looking for a restaurant, bar, store, etc. it is better for the user to list locations, whereas searches for service providers may actually result in more clickthroughs to websites for more information and thus provide you with a blended result?

    Obviously just speculation, but I know in my experience we see more blended results than not and when they aren’t blended it’s usually because there are no local results, even for localized queries.

  • For the 1st few months after the Oct 27th change I was certain it was based on search volume.

    “Core” keywords pulled a blended and longer tail or less popular search terms pulled a pack. (Fairly consistently.)

    Then there was a period right after that where I noticed City KW would show one algo and KW City would show the other. Didn’t make sense at all from a user viewpoint because I would think both searches show equal local intent if the city is in there. Then that changed and searches became consistent regardless of KW order.

    I’ve seen a gradual change where G is showing more and more packs EVEN in large cities for popular core keywords like Dentist and lawyer.

    I also THINK I may be seeing a change in the “pack” algo. Seems it isn’t as strongly tied to # of reviews and citations. BUT that’s just a quick observation, have not had time to dig in and analyze. Anyone else notice this???

    Recently it seems to be pretty helter skelter as far as which layout she’ll serve up and it’s hard to say what she’d doing. Testing I guess because it’s too random to make sense. OR maybe she does it just to see how many SEOs she can drive to the nut house. 😉

  • Hey everyone. Thanks for the awesome feedback and contributions! So many great ideas and theories on this topic.

    I plan to get to the bottom of this 7-pack vs blended mystery. I am currently running a huge data collection study. 77 keywords (many different local industries) in 23 different cities (large and small) using 15 different browsers (including mobile browsers). I’m going to collect the data every day until I feel like I have enough to understand what’s going on, and then analyze it. I’ll post the data and my conclusions to the blog once I’ve had a chance to look at it.

    One other theory that just occurred to me, and which no one mentioned explicitly, is that it could be a simple reaction to ad click through. I wouldn’t be surprised if Google is split testing and adjusting automatically based on ad clicks. If one version happens to get more ad clicks, then they start showing that version more often in the split test. It’s in constant flux depending on which variation tends to be making Google the most money at that specific moment in time. Makes sense to me. If I was Google, this is what I would do.

    @Nyagoslav: I think you raise a really interesting point about it being tied to Google Places data updates. That would make some sense. I’m not certain about it being related to the relative strength of local search factors though. I’ve seen many search terms such as ‘chicago lawyers’ vs ‘chicago lawyer’ where one returns 7-pack and the other returns blended. Presumably, the list of sites relevant for those terms would be the same. If you think about on-site optimization though, it *IS** quite likely that the website could very well be optimized for the singlular version and not the plural version, so, I think you may be on to something with your theory.

    @Mary & Chris: That’s a really good point. That would certainly explain why I saw it switch from 7-pack to blended on repeated searches. I think I’ll try hitting it again a few times anytime I see a 7-pack to see if I can duplicate.

    @Mike: I can find examples that contradict the query volume concept, but that doesn’t mean much. As others have mentioned, the results seem to flip flop between 7-pack and blended even down to the hour, so it could very well still be query volume at that particular moment. I think you’re definitely on to something with your idea about the type of business. My data collection study should reveal some insights around this.

    @Matt: Great insight about the reason they introduced the blended algo! That makes good sense to me. I’m starting to agree with you about query volume and search intent, but only when combined with other factors in order to serve up the variation that results in the highest number of ad clicks.

    @Gary: I *will* get to the bottom of this. I’m on a mission now!

    @Dev: Yeah, I was thinking about the onsite signals and how they might play into this. Once I have the data from my study, I’ll use it to take a closer look at the links on the 7-pack ranking sites.

    @Zunaira: That’s really interesting. It helps to strengthen the query volume theory. If the 7-pack or blended is shown based on query volume within a short time frame, like a number of hours, then that would explain why you get 7-pack in the morning and blended i the evening in industries where people tend to search for the service in the evenings.

    @Linda: Reading your comment really made me think that my theory about ad click through might be on track. At the beginning, they didn’t have click through data, so the split tests were fairly evenly distributed between queries based on some criteria they set. As they collected more data and saw what worked and what didn’t they continued to to throw manual tweaks in while the system was automatically learning and responding. Now, things are pretty much on auto-pilot and the system is automatically reacting to query volume, search intent, and click through behaviour in some kind of complicated and beautiful symphony that maximizes Google’s profit. So basically, all the different theories are probably partially correct. My guess is that it’s a combination of all of the theories into a complex, auto learning, auto adjusting, split testing algorithm designed to maximize Google’s profit. Google certainly has the means and the motive to do exactly that.

    Man, trying to decipher how and why Google does what is does is so fun! I love being an SEO.

    Thanks again for the great discussion everyone.

  • I think you (as well as others here) are in the target zone, Darren. It seems logical to assume that the fluctuations we are seeing are the result of a yet another complex algo which attempts to “think” for itself – driven by a single ultimate goal – profitability. If the underlying premise of providing ultimate user experience is met – monetizing the model is facilitated.

    I will looking forward to your findings 🙂

  • Really interesting comments. I seem to recall in Mike B’s blog there was a comment from a googler referencing that change/experimentation on an issue similar to this one is the google norm.

    There may be some algo behind the two types of change….but there could be a randomness also to assist in their assessment of data.

    and thar is my $0.02

  • I ran a number of tests for different categories of service companies in the same town/city in the UK. The majority of times when a 7-pack was shown it included at least one entry for a Place which didn’t have a website. Do you think this could be a factor in the decision?

  • Given that it is Google we know that it is 1)algo driven and 2)by intent.

    One likely reason it has been so difficult to pin down the cause of why one display shows over another is that there are multiple reasons driving which display will show. And those reasons, or at least variations of them are changing over time.

    This will make it very difficult to analyze even with moderate sample sizes.

  • @Blaine: I think that’s just a result of the 7-pack, not a factor. 7-packs seem to use a different algorithm still, so businesses without websites can rank in the 7-pack, whereas they can’t in blended. I should hopefully know more when I analyze the data we get back from our study.

    @Mike: I think you’re right. I think it’s possible that we’re going to collect a ton of data, and all it’s going to do is confuse us more. We’ll see.

  • Exciting! We will see. can’t wait for the results. I really like this theory based on ad profit. interesting. I am still not really certain how you will determine if that is one of the factors that determines 7-pack or blended….?

  • I’ve had tons of situations where a small business I was working with caused it to change from the 7-pack to the blended. I found it happened when I had a client with a website that had a really strong organic ranking. I also work with a ton of businesses that have more than one domain (not smart, but it’s usually because they are franchises of a big business). So we’ll be using one domain and it shows as a 7-pack. Then I’ll notice that their other domain is actually ranking organically down below. We switch their Places page to that domain and the results change to blended.

  • Split testing makes the most sense here but since the anti-trust law passed on July 25th, Google places has been all over the place! 🙂

    I guess it’s a major issue when you’re forced to remove many reviews and citations.

    Based on that to me it only makes sense that they are split testing everything to see which new version gets better conversions… then again your theory about thresholds is an interesting one to explore.

    Anyways just my $0.02. Great article BTW!

  • Re-reading through some of the posts mentioned here the weighted theory also makes a lot of sense when you think how a search engine operates. Low volumes give the quickest and cheapest result for them to deliver while higher volume queries take a blended approach.

  • Darren, I think your idea about ads is spot on! G must please the searcher first to be able to get repeat users. That inevitably means more clicks on ads, which means revenue for G.
    However, intuitively I would think CTRs on ads would be higher on pages with packs. Less info listed would cause ad copy to be more attractive.
    I could see less CTR on ads with blended results as the searcher can glean more info from the organic results faster.
    But what do I know? However, if this is true, it would be interesting to see if the trend continues towards blended.
    Another theory I have, is that the pack algo is much more easy to understand and manipulate. Once the SEO community “cracked the code”, it was time for G to start implementing complexity to maintain user integrity by including other traditional SEO signals.
    Ultimately it will be the SMBs who optimize all areas of their biz (Places, website SEO, Social networks) that provide the best user experience who will win the ranking wars.

  • Darren first off great site and blog subject. This is something i’ve been testing on admittedly a much smaller scale recently and seem to be drawing blanks so far. I’ll be interested to see the results of your data collection test. Since the revised places algorithm I was certain the organic ranking (together with reasonable citations) was the key to getting a listing “merged” but I still haven’t witnessed any hard and fast results to back this up. With the time issue involved between tweaking place pages and seeing actual consequences in the rankings it makes it all the more difficult to test one method against the other. I’ve recently done some on and off page seo for a client to get the ranking #2 for a geo keyword and set up a place page which surprisingly hadn’t already been “scraped” by Google considering the amount of NAP citations they have. As a test I’ve collected all citation urls I could find, bookmarked them and created a couple of rss feeds and submitted to a few rss directories. I’ll report back if I find this has much affect in bumping the place page and moreover merging with the (already high ranking) organic listing.

  • i am also reading this and i am a customer of darren, this is interesting to read because it happened with me sometimes, i would through a query like security companies london and i would find pack and i would search after another minutes and it would change the whole result and take me out to of results.

  • I tend to swing towards it being down to user intent and the confidence they have that the results they are providing are as good as they can be for a given search query.

    We sometimes see a single localised listing up top, then a few organic results, then maybe 3 or 4 more local results then more organic. The complicated and vague you can make the query the more they try to cover the spread with more variance (videos, images, news etc).

    If they believe that 50% of these searchers are looking for a local business then dedicate 50% of the screen to these results. If they think 25% are looking for images and 25% news then show some of those results and then throw in a few organic to really cover all bases.

    Beyond that I believe personalisation, split testing and general updates are having some impact here.

    An interesting thing we see is businesses with a heavy spam footprint don’t tend to localise. Even website design in the UK, possibly due to all the noise from footer links and the like (? No good quality signals or just other algorithmic noise ?)

    Marcus

  • To further complicate this issue… this past week I’ve been seeing a client’s listing in the 7 pack (root domain) ‘and’ in the organic/blended results with the root domain. This is unusual, as prior to last week I’d only seen the organic/blended search results show a ‘deep’ page at the same time the client showed up in the 7 pack with the root domain.
    In other words;
    this past week I’ve seen:
    7 pack listing widget(dot)com ‘and’ organic/blended widget(dot)com
    prior it was always:
    7 pack listing widget(dot)com ‘and’ organic/blended widget(dot)com/deep-page
    or:
    7 pack listing widget(dot)com/deep-page ‘and’ organic/blended widget(dot)com

  • They knew it was you and they’re populating a specific result for your searches in a last-ditch effort to thwart your WW2-level code cracking efforts on their complex and highly prized search algorithms.

Join the Discussion

Your email address will not be published. Required fields are marked *