This is pretty interesting for those of us devising strategies around eBay.
Panda 4.0, Payday Loan 2.0 & eBay’s Very Bad Day
After a period of relative quiet, MozCast detected a major “temperature” spike in Google’s algorithm at some point on Monday, May 19th. This occurred after some historic lows, including the 3rd coldest day on record (May 11th).
Tuesday afternoon, Google confirmed two updates, Panda 4.0 and Payday Loan 2.0. Matt Cutts tweeted the Panda 4.0 announcement:
Less than an hour earlier, Search Engine Land confirmed the Payday Loan 2.0 update. This ended a weekend of wild speculation (including many predictions of a Penguin update), but didn’t leave us with many details about the timeframe or the impact.
Which update was which?
For the moment, we’re going to have to speculate a bit. If the latest iteration of the Payday Loan update is like the first, it hit hard but fairly narrowly. Google laser-targeted some very spammy verticals with Payday Loan 1.0 (back on June 11, 2013), but the overall impact was moderate. That update was also very query-specific. My gut reaction is that it was unlikely that the May 19th update was Payday Loan 2.0 – that update was probably smaller and rolled out over the weekend (possibly May 16th). There was heavy flux around a few potentially spammy queries on May 16th, including “mortgage rate trends” and “cheap apartments”, but competitive queries tend to change frequently, so the evidence is unclear.
Google’s numbering scheme suggests that Panda 4.0 is a major update, which probably means that it is both an algorithmic update and a data refresh. This typically means substantial rankings flux, and I think that’s much more likely connected to what we’re seeing on May 19th. While Matt’s tweet implies a roll-out on May 20th, most Panda updates over the past year have been multi-day roll-outs. We should know more in the next few days.
What happened to eBay?
Digging into the May 19th data (and before Google confirmed anything), I noticed that a few keywords seemed to show losses for eBay, and the main eBay sub-domain fell completely out of the ” Big 10″ (our metric of the ten domains with the most “real estate” in the top 10). Sites shift, and nothing on the level of a keyword means much, so I took a look at the historical eBay data. This is eBay’s share of top 10 rankings for the past week across the MozCast 10K (approximately 94,000 URLs, since not all page-1 SERPs have ten results):
Over the course of about three days, eBay fell from #6 in our Big 10 to #25. Change is the norm for Google’s SERPs, but this particular change is clearly out of place, historically speaking. eBay has been #6 in our Big 10 since March 1st, and prior to that primarily competed with Twitter.com for either the #6 or #7 place. The drop to #25 is very large. Overall, eBay has gone from right at 1% of the URLs in our data set down to 0.28%, dropping more than two-thirds of the ranking real-estate they previously held.
It is entirely possible that this is temporary, and it’s not my intention to “out” eBay – I have no idea if they’ve done anything that merits major ranking changes. This could be a technical issue or a mistake on Google’s part. It’s also worth noting that these results only track the main eBay sub-domain (www.ebay.com), not other ranking sub-domains, including popular.ebay.com.
What exactly did eBay lose?
Looking just at the day-over-day change from May 19-20, I dug into the keywords that eBay lost out on, hoping to find some clues about the broader Google updates. The vast majority of losses were where eBay had one top 10 ranking and then fell out of the top 10. In three cases, eBay lost two top 10 rankings for a single keyword phrase. Those phrases were:
Here’s what the top 10 looked like for that first phrase (sub-domain only) on May 19th:
eBay held the #7 and #8 spots. Here’s the top 10 for the next morning, May 20th:
It’s interesting to note that both eBay losses here were category pages, not specific products. Here’s one example (from this eBay URL):
For the other two keywords where eBay lost two positions in the top 10, the lost URLs were also category or sub-category pages (not individual auction listings). The remaining losses were either situations where eBay went from two listings to one or one to zero.
Here are the top 25 keywords where eBay lost one top 10 ranking position, ordered by their MozCast temperature:
It’s very hard to interpret individual keyword changes, but, not surprisingly, many of these phrases seem to be products and product categories, and some are fairly competitive. Most of these drops seem to be from lower positions in the top 10 – I was unable to find a case where eBay lost a #1 ranking day-over-day.
In one case, it appears that both “www.ebay.com” and “popular.ebay.com” lost out. Here are the top 10 sub-domains for May 19th for the query “hooked on phonics”:
…and here’s the same SERP the morning of May 20th:
One page on “popular.ebay.com” kept its spot (this category page), but two narrower category pages lost out. In this particular example, Amazon picked up a top 10 spot, although their highest position dropped. Both Amazon URLs were for specific products, although it’s important not to generalize too much from one example.
What does it mean for you?
I’m sorry to say that it’s probably too soon to tell. We’re hearing reports of big losses and gains, which is the norm for any major update – for every winner, there’s a loser. If Google is to be believed, we’re looking at two sizable updates in the span of a long weekend. It’s possible we’ll see even more changes before the US holiday weekend (Memorial Day), so I’d strongly suggest keeping your eyes open.
Update (May 21st – 9:30AM)
Good follow-up post from Rishi Lakhani about eBay’s internal linking structure. Digging deeper, it looks like all of the URLs of the form “ebay.com/bhp” have disappeared from the rankings, at least within or data set. We’ve collected another day’s worth of data since the post was written, and the situation hasn’t changed. This could be a manual action on Google’s part, but it’s hard to tell.
Google is now saying that Panda 4.0 impacted 7.5% of English-language queries. Despite Matt’s “…starting today” statement on May 20th, I (and others) strongly believe the Panda 4.0 roll-out may have begun over the weekend, and is connected to the May 19th temperature spike.