Digital

Google Bear case study: poking the massive beast and losing

Google cutely names some of their major algorithms after animals. Some examples include Google Hummingbird (designed to increase speed and accuracy of natural language understanding), Google Panda (aimed at countering on-page spam / thin content sites), Google Penguin (aimed at counteracting link spam) and Google Pigeon (aimed at increasing the ranking of local listings).

To honor this tradition, today I want to write about an observed effect in Google that I’m going to code name “Google Bear”, in particular referencing the phrase in English that one should not “poke the bear”.

High level overview

Recently in my agency Vine Digital we had an interesting experience with one of our clients Bubbahood. Bubbahood used to be known as Me and My Child and had a website meandmychild.com.au. Due to legal reasons they had to rebrand and migrate their website from the domain meandmychild.com.au to bubbahood.com.au.

meandmychild.com.au was doing quite well in organic search over the years. With the usual ups and downs, the general trend since our agency began working on their SEO and content is upwards, with the organic traffic showing the same.

Everything was fine. As of June 2018, according to SEMrush, they ranked top 3 for 108 keywords, and 4-10 for another 92 keywords.

So, when the time came to migrate to bubbahood.com.au, we were quite confident. We’ve done lots of SEO. We know how to migrate a domain and do proper redirections etc. What could possibly go wrong?

That is how Google Bear was born.

The Migration

We executed a flawless technical SEO migration. There’s no doubt in my mind that everything was done properly. We:

  1. Maintained the folder/subfolder structure identically on the source and target domains
  2. Implemented wildcard 301 redirection sets from http://meandmychild.com.au/xyz and https://meandmychild.com.au/xyz to https://bubbahood.com.au/xyz
  3. Ensured that all content, design and assets were identical (it was just a domain migration, the website was almost identical except for the brand name and logos)
  4. Used GSC’s domain migration tool to notify Google of the migration.

Everything was done perfectly.

But then it happened. The site experienced broad ranking decay for a large number of queries. — Well, technically this statement is not exactly correct. A more precise statement would be that Bubbahood did not correctly inherit all of the organic ranking performance of Me and My Child post site migration.

Now, in the beginning, we expected it to take some time for Google to process things. So we checked and double (and triple, and quadruple) checked our work. (We even hired an external SEO to check it again to ensure we weren’t missing anything). No one found anything amiss. Either several seasoned SEO specialists have no idea how to do the absolute basics, or something else happened. Honestly, I think it was the latter.

As you can see by that graph, while there was some growth over time, as of March 2019 (a full 9 months after the site migration) the site’s ranking decayed to 54 positions in the top 3, and 70 in 4-10. With the total number of keywords on the first page now being 124 (compared to the past 200) we can say that the “ranking dropped by 38%” (as a very rough statement — there are other data problems that I won’t go into here).

Which keywords specifically dropped?

Naturally, we dived into the data to try to understand as much as possible which specific keyword / keyword sets had dropped. Perhaps there were logical explanations? It was a good exercise and it did yield some insights, which form the basis of the topic of this article.

S26 related keywords

S26 is the baby formula brand that the website promotes (listings on this s26 baby formula page). It is well-known in Australia and has very good search volume in general. There are a number of example keywords such as “s26 gold” (2400 monthly searches), “s26” (1900 monthly searches), and “s26 baby formula” (390 monthly searches). In general for S26 related keywords Bubbahood experienced only slight decays, often from position 1 to 2 or sometimes 3. However, this is more significant / meaningful than it might appear, as known click-through rates for position 2 are almost half of position 1 — so a drop from position 1 to 2 can half the traffic from that single keyword. Have that happen for a percentage of the total keywords and you can see a noticable drop in traffic. That’s exactly what we observed from our analysis of S26 related keywords.

Content marketing related keywords

One of the original reasons why Me and My Child’s organic traffic grew so well over time was the creation of genuinely useful content for their audience. Naturally, with new mums being the target market, content was created around subjects such as mixed feeding, the baby development tracker / pregnancy tracker, and a bunch of smaller articles on multiple subjects like breastfeeding positions. These articles ranked quite well pre-launch, but we noticed general decay in ranking post launch, especially for these content pieces.

But here’s the mystery — again, why would these suddenly rank lower after a domain name change? Isn’t the content just as valid, just as useful (or not useful) for the user as it was a few months before? We researched the competition and didn’t notice a sudden increase in competitive articles for all content pieces that we were ranking for. It wasn’t the competition rising, it was us falling.

This research and additional data lead us to conclude that it was what I now call the Google Bear. *Drumroll please* — I will now explain what we think happened here.

Google Bear

Google’s index (their private “copy” or “picture”) of the web is massive. While their crawler is constantly refreshing this picture, it is still huge, and it stands to reason that most of it sits “stably” a lot of the time, without much update, especially in what I call “middle earth” (the land of not important, but not unimportant websites, AKA most of us, akin to the “middle class” of society).

It’s also extremely multi-faceted. The index is not just copies of webpages of varying ages, it also includes things like PageRank. The PageRank of a page is calculated by looking at the PageRank of the pages that link to it. Follow that logic through and you get a constantly-evolving map of websites / webpages and their PageRanks.

We have no certain way to know how often Google actually refreshes their calculation of an individual webpage’s PageRank. It could be days, weeks, months or even years. Given that a significant amount of the general web stays “the same” over time, it stands to reason that Google wouldn’t constantly invest resources in re-calculating things like PageRank all the time. Perhaps it follows changes and let’s that have a knock-on recalculation effect, but it doesn’t intentionally go and “refresh” everything all the time — that would take lots of resources for potentially few actual changes.

Enter the concept of Google Bear. The analogy is that Google is a massive bear that has a picture of the web, as well as models on various things (like PageRank, canonicalisation, etc.) that is sitting collectively in their index. The beast is sleeping. It’s not that it’s not active, it’s just that most of the time it’s not reprocessing your specific website in detail. I’ve seen cases where the text of a page was updated months ago but Google still didn’t have it updated. Again, it’s hard work to keep an up-to-date “fresh” copy of a massive web.

BUT, one day you decide to do something that inadvertently “shakes things up” / “rocks the boat”. You migrate your website to another domain.

When you do that, you now force Google to reprocess everything about your website, because it has to process the 301 redirections, index new “novel” pages of the web etc. In the process, it stands to reason that Google has to now allocate a newly recalculated PageRank to each of your webpages.

Then, enter an additional piece of the puzzle; gradual changes to the Me and My Child website over time that might have not been fully processed due to “the bear” sleeping.

Changes over time

The Me and My Child website was also a constantly evolving website. Without going into much detail, several things happened over time with redesigns etc.:

  1. The link structure of the site changed from previous iterations — some content was buried (in particular in this case, the content marketing pieces were slowly given less emphasis in the linking structure of the website)
  2. The architecture of the site changed, pages being moved around etc.

However, through all of these changes the ranking remained relatively unaffected, until the new site was launched. Many design and content layout changes happened over time, but didn’t have an impact, then after launch they did.

Hence the possible explanation: the bear was sleeping, it was ranking this page for query X and this page for query Y, users seemed to be satisfied so “why change anything”, BUT when we migrated the domain suddenly we forced Google to reprocess everything, thereupon their conclusion was that the way they wanted to rank the site is now different.

The Bear was poked, reprocessing was forced, and thus ranking decay occurred.

I have to point out that all of the above is our best speculation. As anyone working in SEO should be honest about, we don’t actually KNOW for sure how Google works or how they are processing things. But it makes sense given the data and context.

Not rocking the boat

The key lesson of course here is the old adage “if it ain’t broke, don’t fix it”. Sometimes if your website is ranking strongly you should be very gentle with making changes. Who knows — if you change things and Google recalculates, it may not come out favourably for you. Sometimes it’s wise to leave well enough alone, or another way of saying that is leave what’s working alone and keep building “something else” to work even better in addition to what’s working.

Which leads me to my final point:

What can you do?

Situations like this of course lead a pragmatic person to wonder what they can do. Unfortunately in cases like this, you only have one (general) tool in the toolbox: keep improving the website. The data we reviewed suggested in particular that the content marketing pieces could use improvement, so that’s what we’re doing; trying to make stuff better!

Let’s see where it goes from here!

Shawn 21-03-2019

Leave a Reply

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