On the morning of Friday, November 15th, we woke up to a substantial one-day temperature spike on MozCast. Digging in, there were no signs of a glitch, and it seemed to hit across multiple IPs. The 30-day history looks like this:
Webmaster chatter seemed normal and Google has not confirmed an update, but soon other major flux-tracking tools showed one-day spikes. Here’s the data from SERPmetrics:
Both SERPmetrics and SERPs.com (that graph is a bit less clear, due to an unusually low-flux day earlier in the month) show the spike on November 15th, but one-day shifts are common due to measurement differences in the three tools.
Did big sites win big?
The first thing I dig into when we see a temperature spike is a set of secondary metrics that look at large-scale trends across the data set. That morning showed a solid jump in the “Big 10,” which simply represents the percentage of total search results in the set occupied by the top 10 domains (for that day):
The one day jump from 15.39% to 15.89% represents a 3.2% relative increase – it may not seem like a huge amount, but it’s historically unusual. Wikipedia, Amazon, and eBay all had one-day gains in the 3-5% range.
Unfortunately, it’s easy to jump to conclusions but much harder to interpret this kind of change. Some algorithm updates might benefit large brands, but it’s just as often the case that an update penalizes low-quality sites, and the big brands simply end up filling the gaps. For example, if the #10 result on a SERP falls out, and the #11 pops up one position to fill that spot, the new #10 is more likely to be a big site with a large Google footprint than a small site.
Our larger data set (not currently public) set shows a similar trend. All I can say with certainty is: (1) this was a historically unusually one-day change, and (2) the “Big 10” metric is at now at a historical high (going back to April 2012). I have no reliable clues about the causality and what specifically changed to cause this increase.
What did Wikipedia win?
Since digging into high-temperature keywords didn’t reveal any clear patterns, I thought it might be interesting to see where a big winner (like Wikipedia) picked up top 10 listings. In most of the cases I saw, the big domains didn’t gain prime real estate, but simply picked up a top 10 result because another site fell out. For example, here are the top 10 on November 14th for “famous footwear store hours” (domains only):
- FamousFootwear.com
- FamousFootwear.com
- FamousFootwear.com
- FamousFootwear.com
- FamousFootwear.com
- FamousFootwear.com
- FamousFootwear.com
- MyStore411.com
- Wiki.Answers.com
Clear, this SERP was dominated that day by the main brand’s site (in this case, individual store locations). On November 15th, though, it appears there was a shuffle in domain crowding:
- FamousFootwear.com
- FamousFootwear.com
- FamousFootwear.com
- FamousFootwear.com
- MyStore411.com
- Wiki.Answers.com
- OutletLocation.com
- Indeed.com
- Wikipedia.org
- Yelp.com
The main brand’s site dropped from eight results to four, and Wikipedia simply picked up one of the newly opened spots. This domain crowding/diversity pattern didn’t seem to hold up across the data set, but it does appear that the gains by big domains were primarily due to losses higher in the SERPs. In other words, big domains like Wikipedia and Amazon only picked up top 10 rankings because someone else fell out.
Was there a glitch?
Something else happened on November 14th that was a bit odd. I informally polled my Twitter followers about that day and got the following bit of information from Galen Ward:
Coincidentally, I had just been in the Moz Google Webmaster Tools account that morning and happened upon this (I didn’t put two and two together until Galen’s tweet):
I didn’t think much of it at the time (temporary glitches happen), but it seems that multiple webmasters and SEOs got the same error on the same day. Is it possible that a bug on Google’s end could cause large-scale ranking fluctuations? It depends a lot on the scope and nature of the bug. Last April, a Google bug caused a number of domains to be misclassified as parked, and the impact was large enough to cause noticeable ranking changes.
If this was simply an unexpected side effect of a bug, though, we’d expect a reversal. The temperature the next day or soon after would spike again, and the secondary metrics, like the Big 10 increase, would settle back to their former values. In this case, we’ve seen no such reversal.
Is Andy Kaufman alive?
When it comes to daily ranking changes, separating the signal from the noise is incredibly difficult. The morning of November 15th, we captured a change that illustrates just how dynamic Google has become (and is something I’ve wanted to capture in the wild for a while).
Around November 13th, TMZ broke a story that a woman claiming to be Andy Kaufman’s daughter said that her father was still alive. Multiple news sources picked up on this story on November 14th. Early that morning, we captured the first page of results for “kaufman”, which were as follows:
- IMDB (Charlie Kaufman)
- Wikipedia (Kaufman, TX)
- Wikipedia (Andy Kaufman)
- RobertKaufman.com
- KaufmanCo.com
- KaufmanCounty.net
- KaufmanTX.org
- Kauffman.org
- Fandango.com (Kaufman Astoria Cinemas)
Google was viewing a search for “kaufman” as informational and generic, returning results for Andy Kaufman, Charlie Kaufman, cities named Kaufman, etc. A disambiguation box on the SERP even makes it clear that Google has trouble interpreting the query.
After the story about Andy Kaufman broke, the SERP changed dramatically:
- CNN
- IMDB (Charlie Kaufman)
- CNN
- Wikipedia (Kaufman, TX)
- Fox News
- KaufmanCounty.net
- RobertKaufman.com
- KaufmanCo.com
- US Today
- NY Daily News
Where there were no news-related organic results before, news articles now accounted for half of the top ten, including the #1 and #3 spots. You may have heard the term “QDF” (Query Deserve Freshness) in the SEO world. What’s interesting here is that QDF is not something that’s just on or off for any particular query. A query that was relatively static transformed overnight because of new information. In other words, Google decided in real-time that this informational query was now a news query, simply based on new data and content.
Is this the cause of the overall flux? No – it’s very unlikely that a single event could move the needle. Even an event like 9/11, that had a huge impact on many people, is only going to be relevant to a small percentage of queries. Events like these simply go to show how dynamic any given query can be on any given day. In a case like this, the query isn’t even historically high flux – it transformed overnight, and that transformation had nothing to do with algorithm updates.
So, what happened?
If it seems like I’m stalling, then, well – hey, is that Elvis?! One of the difficulties of retroactively explaining rankings fluctuations is that we typically can only look at the results themselves. This essentially means that we’re measuring positions, position changes, and characteristics of the domains and URLs. This makes it easy to measure something like domain diversity but very difficult to profile something like a Penguin update, where the changes are due to characteristics of the individual sites and their link profiles.
We’re also creeping into the holiday season – we’ve already seen a pattern of above average flux in the weekend before Thanksgiving. As we get into Black Friday, commercial SERPs naturally fluctuate, and it’s hard to separate what Google is doing from changes due to competition and seasonality.
Whatever happened on November 14-15, it doesn’t appear to have rolled back. The one-day spike is similar to a more traditional algorithm update, but that’s about the best we have for now. If anyone has seen additional clues or has any follow-up on the DNS errors in Google Webmaster Tools, please leave a comment.
