Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant Flux

Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant Flux

Alternate title: "What the Flux? Why the Google Pigeon Algo Keeps Changing and is So Tough to Figure Out"



Google "Pigeon" flew the coup on 7/24, but this bird has still not landed! If you've been following the "Pigeon" news or if you do ranking checking at all in the US, you know this crazy Pigeon algo is still not settled in and changes every time you turn around. Major flux continues and I'll share some reasons why.

As I was working on the new 2014 Local Search Ranking Factors for David Mihm, which is due to be published in September, it occurred to me Pigeon is one month old. The LSRF timing is really unfortunate, since the algo hasn't settled down enough for any of the experts working on the LSRF to even begin to figure it out.

We've been getting reports at the Google forum from users saying their ranking is rotating back and forth every day. OR they tell us they used to rank B, but dropped out of the pack. We check and they are still B.

I have a new discovery that helps make sense of the flux and could help you, if you are diagnosing rank drops during this uncertain time.

I've had a Pigeon theory for awhile now. I've floated bits and pieces out there, but wanted to try to lay it all out in one place. And now I have PROOF my theory was correct!

After doing lots of Pigeon analysis, I've been convinced for awhile now that Google is testing 2 -3 versions of Pigeon on different datacenters and is rotating results. (Likely for AB testing.)

I say this in part, because all the Google datacenter IPs I can find like THIS ONE (I've tested tons more) are ALL still showing the old pre-Pigeon results. (Check it out at that link and compare to the Pre-Pigeon shot below. Even though many of the results are the same - David H. Lee and Matt Brown-Ruegg only ranked PRE-Pigeon, never POST-Pigeon.)

Classic Google maps is showing the old ranking order too, because it’s still using an old datacenter so that's why results don't match Google search. (David H. Lee and Matt Brown-Ruegg still rank in classic, same order as the 6/22 Pre-Pigeon shot below.)

Plus many of the pack tracking tools I use are still showing the old algo ranking order as well. That's because they are all using older datacenter or maybe I should say non Pigeon DCs.

So anyway, rotating different versions of the new algo on a couple specific new datacenters would account for the crazy constant flux we are seeing. Depends on which datacenter you hit. I repeat - the Pigeon has not landed!

Well yesterday I discovered proof that my theory is correct and I want to show you what I discovered via screenshots.

I've been sharing rank tracking screenshots for "Seattle Chiropractor" since Pigeon hit.
See: Tracking Google Local Pigeon Flux Over Time

Yesterday when the results changed radically again, I dug deeper and will show you what's up. Skipping a bunch of in-between ranking screenshots which you can see at the link above and only showing you the Pre-Pigeon results and 8/23 and 8/26 results.



8/23 - just showing the one to contrast and show the difference post-Pigeon, although as you can see in the Pigeon Flux link above, it's changed daily.



OK now this is where it gets interesting...

Yesterday, like I said, I noticed the pack changed pretty drastically again.



Some significant shifts. (I note them all in the Ranking Flux Over Time post.)

I checked Chrome - results were totally different - normally FF & Chrome match pretty well.

So went back to Firefox in a new window, searched again to compare browser results and saw this.



So I had 2 different results, both using FF. (And double checked, both with location set to Seattle.)

So next I did what any good geek would do and hit hard refresh. And sure enough - not every time, but quite often, I could get the results to rotate between the 2 datacenters/algos.

HERE IS A SIDE BY SIDE COMPARISON​

AND that 2nd algo is the one that Chrome was showing too.

So that crazy bird is flipping back and forth testing! She's definitely rotating datacenters!
Don't like what you are seeing? Hit Ctrl F5, do a hard refresh and check the other version.

Note: does not always work. Sometimes you have do it several times before it will switch over. And I think if the results don't change, then maybe that particular pack does not have the factors that the different versions of this algo affects. I tested a bunch of different results. Some changed and some did not.

No way to know which datacenter and which version of the algo is going to win out. But it's interesting to see and compare the different versions to try to get a preview of what may be coming!

GOOGLE MAPS MISMATCH

I mentioned above that Classic Google maps is showing the old ranking order too, because it’s using a different datacenter than Google search.

FYI new maps WAS on the old datacenter but was recently upgraded. So now new maps correlates with the new Pigeon ranking order, but Classsic does not. Which means Classic and new maps ranking order does not match, which is causing lots of problems here and with users at the Google forum.

SEE NEXT POST FOR DETAIL ABOUT MAP RANKING ISSUES & INCONSISTENCIES.
(Splitting to new post, so this post does not get too long.)

Now you know when you are troubleshooting, doing rank tracking OR trying to reverse engineer this crazy Pigeon algo - it totally depends which browser you are using and which version of maps you are checking. Because it depends on which datacenter you hit.

Many tracking tools are still hitting the old datacenter and showing the old results.

Non US Google search like: Google Germany is on the old datacenter too. (So you can use that or Classic to see/compare the old results.)

So anyway everything is just all jumbled up and it does not make a good user experience for anyone. And it makes it super hard for people like me to try to figure this algo out.

IN SUMMARY

.:. Google Search - rotating datacenters and ranking order
.:. New Google maps - same as above so normally mirrors search results
.:. Classic Maps - old pre-Pigeon results
.:. Many tracking programs - old pre-Pigeon results
.:. Checking US results from - google.ca or .de or other countries - old pre-Pigeon results

So if you don't like the results you see, or if you are seeing something different than your client sees, above may be the reason. Just blink or switch browsers or switch maps or hit hard refresh and you may see something totally different.


Miriam Ellis from Moz asked me to weigh in on an article she's doing there. She asks top experts what their best Pigeon advice is. Here's an excerpt of what I told her...

My advice? The Pigeon has flown the coup, but has not landed yet! There is too much flux for any of us to make sense of yet. All I can say is that hopefully by the time this bird is tested and/or trained, it will settle into something equitable for users and business owners alike.

But no solid advice can be given now, except to keep working on best practice strategies, which will always help in the end.

Tons more info about Pigeon at our #Pigeon Hashtag.

What do you think about all this???

Note: This is all theory based on my limited testing, so feel free to shoot holes. That's how we all learn!

(Next post covers map inconsistencies in more detail)

<meta property="og:type" content="article"><meta property="og:title" content=""><meta property="og:description" content="Google Pigeon flew the coup on 7/24, but this bird has still not landed! Major flux continues and I'll share some reasons why.">
<meta property="og:image" content="">

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

So starting a new post to cover the Google maps inconsistencies in more detail.

In the previous post I mentioned we've been getting reports in the Google forum from users saying their ranking is rotating back and forth every day. OR they tell us they rank in one position and we check and their ranking is totally different. So the previous post explains why some of those issues are happening.

But we are also getting complaints from users that are confused because Classic and New Google maps are showing different results.

And it's not just ranking issues that are affected by the inconsistencies. Inconsistencies like markers in different places or directions being off and other important issues like closed status are coming up differently between the 2 versions of maps.

I mentioned in previous post, that Classic Google maps is showing the old ranking order too, because it’s using a different datacenter than Google search.

New maps WAS on the old datacenter but was recently upgraded. So now new maps mirrors Google search. (The new Pigeon results.)

Therefore Classic and new maps ranking order does not match and that's causing problems

See discussion here at our forum about some of the issues.

Below I'll show that Classic maps is still using the old algo (sitting on an old datacenter).

CLASSIC MAPS - SHOWING PRE-PIGEON RESULTS​

Here is CLASSIC MAPS From 8/23 (Same day as the 2nd screenshot from the top in previous post.)




So above you can see that the Classic maps does not correlate with the 8/23 SERPs screenshot above, but DOES correlate with the Pre-Pigeon results in the top screenshot. (And again the telltale sign of the old algo - David H. Lee and Matt Brown-Ruegg only ranked PRE-Pigeon, never POST-Pigeon.)

The fact that new maps better mirrors the 7 pack leads me to believe, that the datacenter we hit when checking new maps has been updated to Pigeon (and from what I can tell, just recently) but classic maps is still using the old datacenter.

There have been constant changes, sometimes multiple times per day, ever since Pigeon flew the coup. But ranking order has never been similar to the pre-Pigeon result in screen shot #1 In the 1st post). And the Classic maps results have stayed consistent, showing pre-Pigeon ranking order.

So Classic is still sitting on an older datacenter. You can also compare Classic maps ranking order with Google Germany which is on the old datacenter too.

So if a client tells you they dropped in maps, better check different maps and different browsers and hit hard refresh to try to see what they are seeing.

I brought this up to Google and said when results vary so much between Google's own products, it's hard for users to trust the results or know which is accurate.

If you are having problems specifically due to the difference between Classic and new maps, please comment in this post. I've already alerted Google to that post.

But general discussion about all the crazy inconsistencies, feel free to fire away here.

What do you think about the rotation theory?
What kinds of crazy results are you seeing?

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

WHEW!

I've known about all of this for awhile, but you can see why took me time to pull this all together to try to explain. Sorry it's so long. Hopefully makes sense.

Should have made a video, would have been so much easier to show/explain.

Sure wish I could just do a brain dump for you guys every time I discover something!

Anyway I've edited those 2 posts about a thousand times, and then finally called it good enough.
Scuse any remaining typos and let me know if anything is unclear.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Thanks so much for the post Linda! I have been noticing inconsistencies between pack results and classic maps results for a while now but could not figure out the cause. This really helped to shed some light on a very frustrating situation!

I also noticed that ranking sites, like Bright Local for example, seem to be showing pre-pigeon results. I am finding it hard to trust most search results since they can be so unpredictable right now.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

@Linda - my findings are consistent with yours so I thought you'd be glad to hear that it's not just you finding these things!

@Nikki - I also am finding our Bright Local reports showing pre-Pigeon results and agree it's making it really difficult to know where clients are actually ranking.

Some points I have noticed as well is that branded searches and category searches are behaving really differently. For category searches, you noted that results are pulling in listings from other cities. For branded searches, I'm noticing the exact opposite - the map is way tighter and listings are getting locked out that belong to that brand.

Also, I'm noticing page age is a bigger factor. These listings that are spam and resurfacing are all old pages that were created ages ago. In traditional SEO age and trust often go hand-in-hand and I'm seeing a lot of that currently.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Thanks so much Nikki and Joy.

It's good to know I'm not the only one seeing these things and together we can brainstorm and at least try to make sense of this mess.

Nikki glad you brought up BrightLocal. I didn't know about them and other major players for sure.

All I know is the little free programs that send me an email when pack shifts and shows me either a screenshot of the pack or simulated pack order - they are all still showing the old algo! Bright too? Wonder if they realize it???

Joy I'm glad you brought up domain age. That's something I've thought of many times when a funky listing ranks that does not seem like it should. But I had not gotten around to checking yet. So that helps confirm my suspicion.

Now if we could figure out a couple more things...

According to SE Land this algo is more about traditional ranking signals, which to me means organic. Soooo...

1) Why are listings ranking high that have NO SITE at all? Just a G+ Local page?

2) Why are listings like Fremont Chiropractic always on top in Pigeon, even though they dropped out of the pack in the old algo and STILL DON'T RANK IN PURE ORGANIC AT ALL?

All of the other listings that rank high in the pack, there is still correlation with the pure organic rankings. But not that one. Does not rank in the 1st pages at AOL at all. So seems like maybe an organic penalty. Does not rank high in pure maps either and is farther north and has no reviews so does not seem to be local factors.

So what ranking signals does that one have? (BTW it always used to be A in the pack for a long time, but for the past year dropped out of the pack and just re-appeared after Pigeon.)

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Linda,

So just to clarify - I mentioned page age (you said domain age) and I meant the age of the Google Places listing, not the website, although I totally agree that domain age is also a factor.

So according to what I'm seeing, listings that have been claimed for let's say 2 years, have a ranking benefit over a page that was claimed or created 6 months ago. I'm also seeing unverified listings all over the place if they have a solid number of reviews or some other factor. But I still think those ones have been "around" for a long time and thus have more of that "trust" factor I was referring to.

So for #1 - I would say listings that have no site that rank are ones that have been in existence for a long period of time. I wish MapMaker went back further than 2012 in the history so I could test that.

For #2 - I would say a combination of the age of the page + the keyword title.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

LOL thanks for correction Joy.

I meant G+ L page age too. Which accounts for the ones that are unclaimed, no site listed and no reviews. It's just longevity of the listing, which means a mature business that's likely been around for a long time.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

what the flux, indeed.

Linda, so much appreciate the work you're doing on this. And glad to have Joy chiming in.

Until we see where this bird lands, I'll keep soothing clients and pouring drinks.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo &amp; the Constant

Thanks Linda and everyone else commenting in this thread. Just listening and following your examples is helpful.

Linda, is there a way to "hard refresh" on a Mac? I tried control f5 (firefox) and it didn't seem to do anything. But there is an "fn" button on my keyboard that seems to refresh the page, but I'm not sure what the difference between refreshing the page and a "hard refresh" is. Can you elaborate a little?

---------- Post Merged at 08:58 PM ---------- Previous Post was at 08:58 PM ----------

Thanks Linda and everyone else commenting in this thread. Just listening and following your examples is helpful.

Linda, is there a way to "hard refresh" on a Mac? I tried control f5 (firefox) and it didn't seem to do anything. But there is an "fn" button on my keyboard that seems to refresh the page, but I'm not sure what the difference between refreshing the page and a "hard refresh" is. Can you elaborate a little?

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Sorry Steven, don't know on a Mac.

On a PC it's Ctl + F5.

But like I said it does not always do it. And then if I get it to switch a couple times and keep trying Ctl + F5 and regular refresh just trying to see what it will do - then it stop, so maybe it's throttled or something.

Plus some SERPs I just could not get to switch so again, I assume that's if there is nothing in that particular pack that varies between the 2 algos. OR it's just kinda random.
OR Google is purposely trying to drive me nuts. Yep, I think that last one is the answer.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Linda Buquet said:

Nikki glad you brought up BrightLocal. I didn't know about them and other major players for sure.

Click to expand...

Yes, this has been throwing off the Bright all month... mostly in the local tools. I'm sure they know... they are pretty Bright guys over there.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Steven -

Command-R will refresh the browser on a Mac.

Use Shift-Command-R for a hard refresh, which will get rid of cached content.

Hope this helps.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Linda you said it best. "Stick to the fundamentals." That is exactly what we are telling clients who are seeing wild swings.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Craziness... keep us updated on what you're seeing. I'll have to do some more extensive testing of my own on this. Great tips for comparing results!

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Nikki Brown said:

Thanks so much for the post Linda! I have been noticing inconsistencies between pack results and classic maps results for a while now but could not figure out the cause. This really helped to shed some light on a very frustrating situation!

I also noticed that ranking sites, like Bright Local for example, seem to be showing pre-pigeon results. I am finding it hard to trust most search results since they can be so unpredictable right now.

Click to expand...

I believe I read in Blumenthal's blog in his comments that both whitespark and bright local rank trackers are not working right now. Sorry I don't have a link, but I don't remember which post it is, and it was in the middle of the comments...

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Wow, great analysis Linda. Great fact finding. I'm very impressed not to mention thankful.

I had just a few thoughts:

"2) Why are listings like Fremont Chiropractic always on top in Pigeon, even though they dropped out of the pack in the old algo and STILL DON'T RANK IN PURE ORGANIC AT ALL?

All of the other listings that rank high in the pack, there is still correlation with the pure organic rankings. But not that one. Does not rank in the 1st pages at AOL at all. So seems like maybe an organic penalty. Does not rank high in pure maps either and is farther north and has no reviews so does not seem to be local factors."

When I took a look at the Seattle Chiropractor market, it was heavily spammed, as you've mentioned before Linda. Many of the ones ranking in the local pack that don't rank in the organic, they have a Penguin penalty. That's why it can seem so confusing sometimes.

Update: Just checked Fremont out of curiosity and yes, they're definitely in a Penguin penalty.

Add on top of that Map Results don't respond to organic penalties. When a Penguin penalty hits, two things happen:

1) Links that were flagged that gave a person a ranking advantage no longer count toward your ranking.
2) Those same flagged links also cause you to be penalized and drop down even further.

It's important to distinguish these two separate actions even though most people don't. When someone gets a penalty removed or "recovers" they wonder why they're not #1 anymore. They were #1 because of the links that eventually were flagged were giving them a ranking boost and now they are not.

Google Maps is like this. Google Maps does take the 1st action above into account (ranking drop due to loss of links) but does not take the 2nd action into account (organic algorithmic penalty).

I can only vouch for that on Penguin although I imagine it's the same with Panda. Also, Panda seems to rarely be site-wide so it would make sense that Panda would act the same way or in an even lesser damaging manner. And I have no clue about how a manual penalty would play into this. I imagine a local business has never triggered that anyway.

"So for #1 - I would say listings that have no site that rank are ones that have been in existence for a long period of time. I wish MapMaker went back further than 2012 in the history so I could test that."

You could use archive.org. You would need the old places URL though because the correct Google+ URL's may have only been around since 2012. That might be too much work so maybe the Yellowpages URL or Superpages URL and that could give you a roundabout figure. You'd have to be mindful of if YP or Superpages has ever changed their URL structure and if so, you could still probably replicate it. I chose those 2 because they've been around the longest. Also, I would imagine Superpages would be the least likely to change their website. It seems ancient.

Hope that helps!

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

More great additions to this thread.

Thanks so much Joshua!

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant

Joshua, meant to say too...

Good to know others are following Seattle Chiro like I do. And I do it specifically because I know there are some aggressive spammy marketers there.

Thanks for investigating Freemont. I thought it must be some kind of penalty but assumed if they were penalized on the organic side they would not rank in blended either. That's the way it worked pre-Pigeon, but Pigeon let's a lot of listings rank that would not have before.

Re: Pigeon Analysis - New Insights about this Crazy Google Local Algo & the Constant


OK for your viewing pleasure. Just refreshed and got 2 new side-by-side comparisons.

The ranking order shifted again - in a drunken flux! (Thanks Justin, told you I'd steal that line!)

Here's 8/26 again - 2 different algos caught at the same time by refreshing browser:



And then here's today - 2 different algos caught at the same time by refreshing browser:



Actually now that I can see them stacked together like that, I guess the shots on the right are pretty similar. Fremont just moved up one, but everything else stayed the same.

But with the 2 on the left, you can see there has been lots of movement - with some previous players out and new ones in.

And interestingly Flourish and Kirsty Smith and are the same practice. I think I may have mentioned somewhere, maybe even this post... I know I just explained it in detail in training, so maybe that's where... anyway... She has her G+ L page set up wrong.

In the past it totally locked her practice out of the pack - but now Pigeon is letting it rank. But then recently sometimes with this rotating algo, the Kirsty listing would knock out the practice. And sometimes they just take turns ranking.

But today in the lower left screen shot they BOTH rank! That's the 1st time I've seen that in this pack.

However someone complained today in another post, about the entire pack in Colo Springs being taken up with 7 Drs from the same practice and according to him they are all bogus listings.

Is anyone besides me getting dizzy yet???

FYI TO FORCE THE SWITCH - I kept doing a hard refresh and it would not switch over.
So then I clicked regular refresh a couple times right in my browser bar and it switched.