Google Webmaster Tools Displaying Incorrect Keyword Positions?

[robzilla]

A couple of days ago, I first noticed that, according to the Top Search Queries report in Webmaster Tools, a site of mine is now ranking on page one for a popular and competitive keyword. There was much rejoicing, obviously. However, so far I haven't received any visitors from it, and when I use a rank checker they never find the site within the top 100.

I know results to differ when I perform searches myself because, among other things, I don't reside in the United States. The rank checkers I tried are hosted within the US and some of them try different data centers, yet all of them tell me the site is "not in top 100."

It's extra peculiar because the keyword ranks at the top of both the Impressions (50% of top 20) and Traffic (20% of top 20) lists on the Top Search Queries page. I made sure I set the Search Type setting to "Web Search" and Location to "(United States) Google.com." The time frame is set to "1 week ago" - when I go further back, the keyword either doesn't appear on the list or it is positioned at 100+.

"Position" is the highest position any page from your site ranked for that query, averaged over the last week. Since our index is dynamic, this may not be the same as the current position of your site for this query.

That's what Webmaster Help has to say about the Top Search Queries [google.com] report. If it had been the highest position a page ranked for at one time, it may have been a glitch and I may have ranked there only for a short amount of time, but here they're talking about a position that is averaged over last week.

Does anyone have any experience with this? Are positions reported in Webmaster Tools indicative of future rankings, or are they known to be unreliable?

[tedster]

Webmaster Tools reports of all kinds are known to contain wrong information at times. This kind of wrong information would be particularly distrubing, but in any big system errors do creep in. The evidence of your own server logs is more dependable.

No, I've never heard anyone mention a future good ranking in their reports before it shows up live - and given the way Google changes their algo I find that quite unlikely, too. This sounds more like a bug.
help
That quote you shared from the Help page is really quite vague, isn't it? As a math major in college, I cringe at the words "highest" and "averaged" appearing in the same sentence.

[robzilla]

Vague, indeed, and contradictory. With other sites and keywords, the positions are usually off as well, but only slightly; these minor changes could be attributed to regional and datacenter differences, or even personal search settings. They often match up with my own statistics, too. This is the first time, however, that the information given simply appears to be wrong.

You're probably right about future rankings. I had my (vain) hopes set on that, for obvious reasons.

[Receptional Andy]

the highest position any page from your site ranked for that query, averaged over the last week

The only way I could read that sentence so that it would reflect data I see in WMT would be that the average is over a time period - so for instance the highest position each day, averaged over 7 days - this way, sites that appear "too high" in WMT are explained by Google's ranking-jumping behaviour - then you only need to be in first place for a single search a day for this to happen.

[robzilla]

then you only need to be in first place for a single search a day for this to happen

If that's the case, the position reported should slowly (or radically) decrease over the next couple of days. So far, it hasn't, but I do see the percentages going down - 25% for impressions, 10% for traffic (compared to the previous 50% and 20%, respectively) [see correction note below]. The "1 week ago" label I also find confusing as, to me, it implies that it's displaying data from 7 days ago, not necessarily from the last 7 days. Unless I take it very literally and replace "ago" with "past" and thus interpret it as the last 7 days gone by, which is probably what they mean.

Correction: I did not have Search Type set to Web Search this time. As soon as I did, the old percentages were back.

[Receptional Andy]

the position reported should slowly (or radically) decrease over the next couple of days

I'm not sure why - the so-called yo-yo effect [webmasterworld.com] results in some URLs ranking well only for short periods over weeks or longer.

That's not to say I think the data (or indeed the documentation) is necessarily accurate - there are marked problems with most of the stats Google makes publicly available, including services like Google trends and the like.

For instance, the geographical targeting options can be pretty skewed - when google.co.uk was showing up in Trends for Websites last week, the biggest region of the UK for visits was Wales, apparently ;)

[robzilla]

I'm not sure why - the so-called yo-yo effect results in some URLs ranking well only for short periods over weeks or longer.

Then I'd expect to be seeing some evidence of traffic from higher rankings in my server logs but, alas, I've found nothing. Even if the site were to show up for only an hour or two, this particular keyword is popular enough to at least send a noticeable number of visitors. Even if it were only 5 visitors, I could easily find those in the logs.

The data is probably accurate in the sense that it's the result of some kind of calculation, but as long as the documentation doesn't explain what exactly is calculated, or presented in the charts, then that data is practically useless or, at least, shouldn't be taken seriously. That's a bit harsh, perhaps, since many other keywords I checked are presented fairly accurately, but what's the point of Top Search Queries charts if I have to check their accuracy all the time, if I can't depend on it?

Aww, shucks, still a junior.

[tedster]

Here's a thought - maybe the WMT report is pulling the position information before some filter is applied to come up with the final rankings. Even though that would certainly be buggy behavior, it might accidentally be showing you that your url COULD rank that well, if only you weren't tripping some kind of filter.

This idea popped into my head mostly because you weren't surprised by seeing a high ranking. You might look at your page to see if it looks "over-optimized" in some way.

[Receptional Andy]

Interesting idea, tedster. That would imply that the "click" and "impression" data was a projection - along the lines of the adwords tool that shows impressions/clicks based on the costs you enter (which also gives some unreliable results).

[ChicagoFan67]

Yes, what Tedster says! I had the exact same thing happen to me. I couldn't get a new page to appear in the serps for many months. It looked like there was some kind of penalty on it - possibly for reciprocal linking or too many affiliate product links. I updated and made some drastic changes to the page and then a few weeks after, GWT reported it at #1 or #2 But it was atleast a good 2 weeks before it finally showed up in the serps.

robzilla, let us know if turns up.

[robzilla]

Not sure how to verify that, tedster, but it's an interesting thought. Indeed, I wasn't wholly surprised as the site, compared to its competitors, all things considered, should, at the very least, rank within the top 20, but never has.

possibly for reciprocal linking or too many affiliate product links

Funny you should mention that, ChicagoFan67. About a month ago I made some changes to the site, a fairly old one, the most significant of which was to undo all link exchanges. Most of these were on the homepage, but some were site-wide. I wanted to get rid of these because it's been a long time since I've participated in link exchanges, have since stepped away from them mostly, and I wanted to clean up the site a bit to prepare for a big overhaul next year.

The site has been optimized, but not to extremes; just basic on-page SEO with which I doubt I've tripped over some filters.

Obviously, from Google's perspective, this was a big change in the site's linking graph (for lack of a better term), but I haven't really seen any big changes in terms of search traffic.

GWT reported it at #1 or #2 But it was atleast a good 2 weeks before it finally showed up in the serps.

Perhaps it's indicative of future rankings after all ;-) Thanks for sharing that. May I ask what kind of changes you made to that particular page?

[tedster]

URLs that show a very large jump in (preliminary) ranking these dayS may not end up actually ranking at the new high position that those preliminary calculations might indicate. This is particularly true for high volume keywords. There seems to be a factor in the algo that puts the breaks on for big ranking jumps.

Sometimes the Yo-Yo effect [webmasterworld.com] can kick in, giving the url a few moments of testing at high rank here and there. There can be other factors that Google checks out after a big jump - such as looking at new backlinks for signs of link buying or other non-"editorial" link growth.

[ChicagoFan67]

Perhaps it's indicative of future rankings after all ;-) Thanks for sharing that. May I ask what kind of changes you made to that particular page?

Yes, that was my thought at the time.

The changes included:

1. extending the title on the page by one word so that it was unique and not the same as the search phrase and every other site trying to compete for that search phrase.

2. I deleted a handful of affiliate product text links. Also, some of the links were doubled up with the inclusion of a thumbnail and had the search phrase in the anchor text. Keyword density may have been part of the problem.

3. The main page and a handful of other pages in the directory all contained a reciprocal link to the same site and all these pages were having difficulty ranking. I had an obligation to keep the links so I took the keyword out of the anchor text and made each link look unique. If reciprocal linking was part of the problem, Google seems to have forgiven me.

4. I rewrote an introductory paragraph on the page which I had previously copy/pasted from another page on my site.

What's different in my case is that I don't recall seeing the search term in the traffic column like you are reporting.

One other thing - I was showing up on Aol search before Google.

[edited by: ChicagoFan67 at 1:56 am (utc) on Oct. 20, 2008]

[g1smd]

The WMT data for site on a .org.uk domain consistently tells me that some page is at #2 for some keyword. That's unlikely as it is a small site and the particular keyword has a lot of competition.

However, every time I go too look at the SERPs for that keyword, it is the .org.au of the same name that is there at #2.

Other than one being called example.org.au and the other being example.org.uk there is *no* connection between the two sites - owner - hosting - location - content - so why this happens is a mystery.

[robzilla]

Quoting a post by tedster in the Yo-Yo Topic [webmasterworld.com]:

[..] one trigger for the Google Yo Yo seems to be a sudden increase in ranking, according to the initial algo calculation. It looks like Google will not stably award a suddenly higher ranking in a competitive SERP.

Do you think it's possible that the removal of so many external links from this site could have brought on an overall boost in PageRank, since more of it would now be flowing through the site and back to the homepage compared to when there were a lot of links to other sites, or that perhaps a previous demotion as a result of the (site-wide) link exchanges is slowly being lifted? As a result of that, the site may have gone up the rankings disproportionally, say from 100+ to top 10, and have hit some sort of trigger..? I'm inclined to think it has something to do with the recent site changes.

ChicagoFan67:

One other thing - I was showing up on Aol search before Google.

But not from the very beginning, if I understand correctly? Search results on AOL seem to be in less of a flux, and perhaps aren't used for experimenting as much as on Google, if at all, so we may not be seeing these effects there.

g1smd: From the surface, then, it looks like they're just not looking any further than the first part of the extension, i.e. .org. Not very Googlesque.

[JackPage]

Here's a thought - maybe the WMT report is pulling the position information before some filter is applied to come up with the final rankings. Even though that would certainly be buggy behavior, it might accidentally be showing you that your url COULD rank that well, if only you weren't tripping some kind of filter.

I've been thinking this myself for awhile and will take a moment to share my own experience.

As I mentioned in a previous topic the queries are showing that I should be ranking rather high for a number of keywords that I'm currently not. The positions Google lists for them makes sense to me - they're unique content pages with links going to them. I can definitely see them being in the listed position.

However, I currently seem to be hit by some sort of filter. It took me out last month for 25 days, removing about 80% of my search traffic overall. The site bounced back afterward for five days, enjoying increased traffic overall, before being hit again by some sort of filter or something that's sent it back into the abyss. Originally, the penalty happened not long after I made some major template changes, so I figured Google was simply reevaluating me and once it found the content hadn't really changed, I'd be back in business. I have no idea what tanked my site again five days later. Total white hat, over a year and a half old. All I can do is hope it comes back after 25 or so days again, a time frame that should be up in about 10 more days.

The interesting thing however is outside of a couple search phrases, I'm getting no more than one hit exactly every 24 hours for almost every other keyword. Many of these keywords are those that WMT's top search queries says I should be ranking for. It's as though once a surfer find one of my pages in Google via a particular term, Google removes it from the SERPs entirely for 24 hours. It's the strangest thing, as these are terms that I should at least be getting a few hits a day on. But no more than zero or one exactly every day just seems odd to say the least.

I have to think that if it weren't for the filter, my rankings would at least be hitting somewhere in those ranges that top search queries is reporting to me.

[robzilla]

As of today, the keyword has gone down both the traffic and impressions lists, in terms of percentages and positioning, indicating that the site now supposedly ranks at the end of page 2 (which it still doesn't). It'll be interesting to see if this continues over the next couple of days and, if so, where it will come to a stop, if at all, and whether that will be at the old position or somewhere in between.

JackPage, have you looked up those single hits in your server logs to see what exact page they were coming from? I'm asking because most website statistics only provide the keyword, not the exact referring page (i.e. the full Google URL with all variables including

start=

).

[robzilla]

Although I am still unable to locate the site in the top 100 search results for this particular keyword, traffic is gradually increasing. The keyword is now sending approximately 8x more visitors than it used to, compared to 5x yesterday, 3x the day before yesterday, and 2x the day before that. 12x tomorrow? We'll see.

In GWT, the keyword is still fairly high up on the Impressions list, but much lower on the Traffic list. The ranking indication seems to have stabilized somewhat, now positioning the site on page 3.

[fred9989]

Looking at my own data in webmaster tools, I'd have to conclude that the query results shown there are "pre-penalty". For about 90% of the data, the results are accurate and represent what I see when I run the searches in Google myself; for the other 10%, the positions on the WMT data are what I would expect if the site was not languishing in the -950 region for those queries (which it is).
While that may seem like an imprecise statement based on my judgement, more importantly, the positions reported in WMT for that 10% of queries actually mirrors closely where the site used to be for those queries before the penalty was applied.
I hope that's clear if not too articulate!
Rod