Back in February I contacted WPengine to ask if we could upgrade to PHP 5.5.
With WPengine (and the way we’ve set up our DNS) this upgrade involved an IP address change in Builtvisible.com’s a-records.
Following this change, I saw a drop in SEO visibility in our UK rankings so, I investigated.
Sure enough, Searchmetrics showed a drop in UK rankings, but curiously, the US data appeared to show a big improvement. It became quickly obvious that the IP change was to blame, a location lookup revealed the problem.
I’d originally set up our hosting to serve from a UK data centre, but when we updated the IP address, I failed to check the new host location. The new IP was located in the USA.
I got back in touch with WPengine who provided me with a new UK IP address, and after changing the location back in the DNS a-records, UK rankings returned.
Given the impactfulness of the IP location change I thought it might be a smart move to repeat the change to see if the ranking flux could be repeated. That’s possible because WPengine keep a routing map of their IP changes, so for now it doesn’t matter which IP address we use, we’ll arrive at the same website.
This brief post shows the change timeline by date and the resulting ranking shifts:
Here are the key dates for changes:
- IP changed to USA on 17th Feb 2015 (188.8.131.52)
- IP returned to UK IP on Feb 28th 2015 (184.108.40.206)
- Repeat Test: settings returned to USA IP on 9th March 2015 (220.127.116.11)
- Repeat Test: returned to UK IP on the 17th March 2015 (18.104.22.168)
This is our own ranking visibility score for the UK search results (click to enlarge):
As you can see there’s a drop immediately after the first change (17th February). The drop is quite significant (we’re tracking 281 keywords that represent a number of important commercial, branded and informational search queries with search volume). Our host IP was returned to the UK IP on Feb 28th. The astonishing thing about the data is how quickly Google responds to this change. The IP change is repeated and a similar behavior is observed.
This is Searchmetrics data for the UK (click to enlarge):
Obviously, we’re still waiting for the 2nd drop in rankings to recover. That update should be the end of this week. As the previous drop and recovery matches our own rank tracking, I’m pretty positive it’ll turn out the same.
This is Searchmetrics data for the USA (click to enlarge):
Rankings went up quite nicely in the US but the drop as we return to the UK IP doesn’t seem as significant. I don’t really know what to think here as the US Searchmetrics update intervals are out of sync with the UK. I suspect we’ll see a drop in US visibility by the end of this week.
Webmaster Tools Search Impressions UK (click to enlarge):
You see a similar drop in overall impressions in the UK between 21st February to the 1st March which matches the timing of the first change. It looks like the same behavior is repeated around the 17th March.
Webmaster Tools Search Impressions US (click to enlarge):
I’m just not sure this is as significant, GWMT reports a 5% improvement in impressions but when we have 886,000 or so impressions in the USA in that period I think we need a bit more data to have any certainty. The IP address itself is located in Texas, and as we don’t have a state level of granularity in our ranking data, it’s very, very hard to see what’s happening here.
I think we see around a 15-20% flux in UK rankings caused by an IP change out of, and back to the UK. Searchmetrics rank tracking for their UK tracked terms matches fluctuations in exactly the same period, although the level of flux will be different as the keywords they track are similar, but not exactly the same as ours.
There was a rankings boost in the Searchmetrics US data set when the IP moved to the USA, but from there it’s a little difficult to unpick what’s happening. My guess is we’ll see an eventual drop in rankings, but another week or so of data would certainly help.
It’s important to note this test was carried out on a “.com” domain with a UK and US IP address. In case it’s not obvious, we use a CDN to serve static JS, CSS and images but the domain itself (and subsequently, the HTML served) resolves via a fixed location dictated by the host IP.
I’ve played with the GWMT International Targeting settings and found setting our “.com” to target the UK improved impressions in the UK but reduced impressions in the US. That’s pretty understandable. I don’t think the change had as powerful effect on UK rankings as the IP change did.
If you’re using a “.co.uk” domain in the UK with a US IP address, the “.co.uk” ccTLD may be a strong enough signal to make an IP change irrelevant. I’m not sure this is the case, as we had no scope to test it.
If you’re operating a multi-regional site, obviously you have href lang and the opportunity to individually target each of your subfolders to a specific location in GWMT, but there may be an additional ranking boost from serving your subfolder content via proxy in the appropriate country.
It’s pretty clear that a UK company using a “.com” or non-UK specific domain should host in the UK. I was lucky enough to be able to test this change quite easily and saw some very interesting behaviors. You should carry out your own tests before making any decisions about your hosting though!
Searchmetrics updated a little sooner this week than I expected it would – here’s the updated visibility for 25th March 2015:
As predicted, rankings have returned.