|Overnight change in keywords that bring G traffic - but same amount of it|
Does anyone know of an update/refresh that happened over this past weekend - early this week while Sandy was trashing NY and NJ? I bring Sandy up because with some 7M customers (say, at least 15-20 M users?) without power it would be a good time to implement updates and such while traffic was light.
Anyhow, I don't use real-time stats but I do compile them on demand using the Vistors software and, comparing stats for 10/27 and 10/31, I can see that almost all of the keywords that bring my Google traffic have changed. The strange thing is that the total amount of traffic from Google is almost the same (adjusted for those several million people still without power here in the NE).
This is a large forum covering a great number of topics (was not built targeting any particular keywords) and I've seen the same set of top keywords bring the most traffic since 2010 - first I started collecting these stats. These "top" keywords were by no means "trophy" - say, only 50-100 visits a month would make them easily stand above the rest. Most KWs are and have always been very long tail and bring 3-10 UVs per month, if even that.
So, I loaded my stats for today and I can see NONE of those top keywords bring any traffic. Judging by the unchanged amount of Google traffic, they have all been replaced by larger amount of different long tails.
Another thing has happened (though I'm not sure it it's essential) when I check the SERPs, almost without exception, Google "forgot" that my site is a forum. The "12 posts - 5 authors - Mar 20, 2008" line is gone from my site's listings (but not from others on the same SERP)
I have to add that there were no significant structural changes on the site since May (one section added) and the content was being added at the normal pace (except for the last couple of days of course)
Does anyone have an idea about what these flips in the set of the KWs signify? I guess, my major concern is whether it's an indication of some issue with the site.
This sounds like the kind of issue some are wrestling with in the traffic shaping and zombie traffic [webmasterworld.com] discussion. And no, after a long time we still don't have any consensus in this area at all. Just some suggestive guesswork and observations.
Maybe your approach to analysis can turn on some light bulbs. Most of the observations so far have been from commercial sites, not forums.
Tedster, you're back! I think you're little ways north from the storm's path but I may be mistaken, anyway, hope you're OK. I thought you may be sitting without power having not seen your posts for a whole day :)
Anyway, getting back to the topic: those aren't really zombies, this traffic is as good as the "normal" was. It's just strange that they all of a sudden decided that the site is more about Y rather than X which they thought it's about for the last 2+ years. And, BTW, there's not that much difference between X and Y as far as the subject matter - just a different set of words.
It may also be simply that I got seriously demoted for the older KWs and the new ones just became more visible. Also, there's always the large amount of visits with KWs not reported, and that will skew the data. I'll try to think of a better way to sort that out.
- Since Google switched to a secure(https) site the incoming keyword is not provided for text searches, are you looking at image data?
- Top pages will often have such a long list of keyword variations that even the best keyword for that page will only represent 10% of the total traffic, if that. Are you recording all keywords for a page and not just the best keyword?
- I've seen odd stats before(haven't we all) in which I would get spurts of traffic in 4 hour intervals for some pages/keywords and then go 4 hours getting it on others instead. It always seemed that I could predict x amount of traffic before the traffic would stop for any given keyword until the next 4 hour period began. I've stopped watching that closely unless a problem arises but this was just before Google switched to secure servers.
- Perhaps most importantly: Keyword research got a serious kick in the pants when Google started showing 95% of pages from one site only on top of any given serp. The reason being your TOP page will be ranked for a search based on the keyword but the rest of the pages Google shows may not even be ranked well for that keyword, they are simply related somehow, yet if a visitor likes the title on the 3rd page shown more it gets the click. In the past that never happened, the 3rd page (or 15th, 38th etc) would never appear for any given keyword because you had a better page.
Good luck getting accurate stats now with the sheer number of tag-along pages getting clicks on keywords they don't rank for, at least not well and never ahead of your better pages, but appear for with those pages.
example serps for keyword "widget", all from same domain
I search for widget, your best page is #1 but I like title in spot #4 more. Your stats pick up #4 as getting a click for that keyword but #4 would NEVER have been there forthat keyword were it not for page #1 and for Google's new way of showing many pages from one site. When you begin getting these multi-page results your stats are going to change.
ie: do you REALLY rank #1 for a keyword or are you ranking #1, #2, #3, #4, #5 etc and the 4th page is getting a higher CTR despite not being your best option for that keyword? Good luck tracking that, especially when your "extra" pages are shown only intermittently.
1script, I've seen very much what you're seeing off and on, and talked about it some in the thread Tedster links (even though I'm not sure it's zombies/shaping, but it seems to be related somehow - maybe).
Sometimes even just for half a day, I'd be ranking for one set of keywords, and then later in the day it would flip to a noticeably different set. My "usual" set is competitive phrases with just a couple of long tail words added on. My "alternate" set was much more long tail, but much more specific/accurate. The bounce rate and other user metrics were much better with the alternate set, but in the end Google settled back on the "usual" set.
Sarge, you are making all great points. I think you are essentially saying that you cannot *really* know which KWs and positions they are coming in on, and I do agree with you. In my case, however, I am not even that concerned about individual KWs - more like trends if you will.
I am only going by results of the unsecured searches - obviously there is no way to know which KW sends a secure one, so that's a huge chunk of my traffic that I cannot analyze right there, and there may very well be some surprises there, too.
By the way, there's *some* info you can still get from secure searches - Google continues to set the "cd=" parameter, which is the SERP position/rank.
So, anyway, all the caveats you list are true but they should be working equally for all KWs, and I just can't see how the SERP layout and consequent user interaction would affect the set of keywords. If the keyword no longer appears in my stats after having been consistently there for years, it should mean that my listing is gone from page #1 - never mind the actual position, title or how enticing the description is.
With all that said, I should report that I have started seeing the usual "top" KWs with pretty much the regular frequency in my stats again this morning. This would make it a 48 hr cycle or just about, so it looks like I'm seeing something similar to what diberry is seeing.
We also enjoy long periods of stable referral patterns. We find that a significant referral shift is followed a week or so later by a general re-ranking for our site, which is not part of a main algo update.
The reason for this seems to be a re-profiling of the site- at the whole site level. In the past, we have seen completely different customer behaviours either side of a shift. Having been through the process on many occassions (and always one in late Sept to early October), we have contingencies to cater for most valid user-intents, but it usually turns out for the better, eventually.
@Shaddows: is the referral shift you're describing continues until the re-ranking? I am seeing the old KWs back in my referrals, so I'm wondering if this fits the pattern you described.
No, we get a temporary shift, then normal traffic (plus zombies- note this is extra traffic, not replacement traffic) then a completely new referral pattern. Previously unloved (but no less worthy) pages get traffic and increased sales, while solid sellers suddenly stop selling completely.
On one scary occasion, a number of our top selling lines dropped 10-15%, but close analysis showed revenue was mostly made up elsewhere. Over time, sales on the top lines returned, but we also kept the new revenue as well. As I said, it can be disruptive but has usually worked out well in the end.
I definitely see a MAJOR traffic change, however, this time I doubt it's "Zombie" related. This looks like another update has hit and this one has hit us hard. Traffic is down since Friday and conversions are non existent. I'm hoping it's temporary, but all the serp changes since early summer have so far stuck. At any one time there are about 4 or 5 people on the site, that is down from the usual 15-20 visitors.
Image heavy blog posts seem to be the going fad.
A new year = new set of keywords (again!)
Just wanted to post an update - the very same thing happened again tonight.
Since I posted this back on Oct. 31, the site had seen the old keywords return gradually and was back on the original set of KWs bringing roughly the same amount of traffic they were for a couple of years before that since mid-November.
Today it looks like another keyword set has been enabled, and it's a completely brand new one, not the ones I saw in October/early Nov. Just like the last time, the total amount of traffic is roughly the same (well, accounting of the Jan 1st hangover) - and just like the last time those KWs I was ranking for for years are nowhere to be found.
What do you think may be causing these KW set switches?