The client had two .com domains. One represented all European countries, while the other domain represented the US and all of the countries globally.
Google was seeing these domains as competing sites.
Each of the domains had lots of country folders, one for each country/language combination. But Google didn’t know which folder to rank in which country.
The whole things was a mess.
Part of the problem was the US site redirected all European traffic to the home page country level folder of the relevant EU site. This was bad for UX in theory because if a UK user clicks on a product page search result from the US site they would find themselves on the home page of the UK site. instead of on a product page.
Therefor they would need to navigate all the way to the product page from there thus being more likely to abandon the journey.
Osprey Europe made an agreement from Osprey US to try and resolve the issue with iThinkMedia’s SEO support and guidance.
The client only wanted to implement hreflangs on their product pages. They said that they planned to do the remainder of the site at a later date which was yet to be determined.
The European site had 32 different country/language combinations while the global site had at least 23.
The challenge would be working with two separate bodies (Osprey US and Osprey EU) and their developers, each with different priorities, objectives development timelines and even time zones.
The developers had a challenge with regard to how they would keep the hreflangs in sync between two completely different sites with separate databases. We spotted that the products had certain EAN numbers that were being used between both countries. We proposed that we use these as a kind of unique ID. The developers agreed that this would work if they were to start including the numbers on the site when new products are being uploaded.
The client’s situation was quite unique because Osprey US & Global ware using IP based redirects, which to some extent couldn’t be switched off.
Another challenge, which became apparent later, was that there is a general lack of information on the Internet on advanced hreflang implementation. Due to the volume of pages we decided it would be easier to manage and better for server performance if we went for the XML hreflang implementation method. But large-scale hreflang implementation using the XML is much less discussed on the internet than the head tag method. This just means if we hit problems we’d needed to figure it out ourselves.
And of course we did hit problems - the initial round of changes just triggered lots of errors in search console. And the non-reciprocating error being identified by WMT for the errors was not correct.
This resulted in a weeding out process where we conducted lots of research, elicited help from other experts and suggested various different changes in an effort to solve the problem. Throughout this process we came across numerous bugs in the development implementation, which we fixed, but nothing seemed to get the herflangs to do what they said on the tin.
Eventually, we came to the conclusion that one of the prerequisites from the US & Global site must be the cause of the problem. This was that the X-default was a URL on their site that redirects based on user IP address. This redirect could not be removed as there was no actual page on this URL. This issue affected every URL set.
The US & Global site’s manager agreed to remove these redirecting pages as the X-default value. In fact we removed X-default entirely. Within days of this change the results kicked in.
We managed to get the job done just in time for the big Christmas rush.
We were measuring rankings on the UK folder on the European site. As a result of our hreflangs, the Google UK rankings increased or gained new position on 486 keywords. We saw a loss in rankings at the same time for just 19 keywords which were not associated with product pages. The massive Improvement was mainly product names that were previously being outranked by the US site.
In the graph above we can see the impact year on year by looking at the uplift in the trend line. This data offers a zoomed in view of a selection of product folders that were affected by the work WE did.
On the bottom-left and bottom-right of the above image you can see the period between when the hreflangs started being implemented, and the end of the year.
It shows a huge 186% increase in sessions based on our sample.
Now let’s look at the impact on revenue.
Our sample of product folders shows a 120% increase in revenue and 100% increase in transactions year on year.
Looking at the specific pages we impacted with hreflangs and looking at how our work affected the main European country folders that we targeted, we can see traffic increases across the board.