Register / Login
Translation of my german speaking SEO-Blog. Many features still missing but quite a lot of postings are already translated. Have fun -- Johannes

Review of Google's Mayday-Update

Google implemented two major changes this year that had an impact on the ranking: Google Caffeine and the Mayday-Update. With the first, Google upgraded their infrastructure to meet the changed requirements for size and speed; this did not incorporate any changes to the ranking-algorithm though. The Mayday-update (which got its name from going live on May 1st) is a different story alltogether: here, Google changed the algorithm to show better different searchresults. The brunt of the changes could be seen in the longtail, which are those keywords that might not get a lot of search volume by themselves but which, summed up, make up the majority of traffic for a number of different types of sites. In the Toolbox, not only do we have our database which gets updated on a weekly basis, but also one that maps the aforementioned longtail. This database consists of more than 10 million keywords and is being updated on a monthly basis. The following analysis refers to this source.

Oftentimes, when Google changed their algorithm over the past few years, it meant that they were putting more weight on the domain-trust, which is the trust that Google has in the quality of a domain. While this might not necessarily increase the quality of the top-10 results, it does decrease the likelihood of there being major discrepancies. To test for this, I took the amount of unique domains in the top-3, top-10 and top-100 both before and after the update and evaluated the results:

Amount of unique Domains
Before MaydayAfter MaydayChange
Top-31.787.0591.776.188-0.7%
Top-103.523.3323.677.257+4,3%
Top-10012.489.20511.712.946-7,3%

Sadly, we don't get a clear picture here: in the top-3 there are next to no changes at all, the top-10 actually sees an increase in unique domains and only the top-100 shows a noticeably decrease of unique domains after the Mayday-update. Maybe Google tweaked their trust-dial with this update, even though it seems that it was more geared towards the lower ranks – it seems they are using different signals on the relevant positions on the first results-page.

The following shows those domains that gained power after the Mayday-update, in absolute keyword-amounts. I tried to limit the list to those domains where the changes are very likely due to Mayday and not due to the overall growth of the domain:

Winners
#DomainBeforeAfterGains
1facebook.com896.2422.067.1891.170.947Domaininfo
2youtube.com2.491.5563.067.692576.136Domaininfo
3wikipedia.org4.095.1984.560.067464.869Domaininfo
4gutefrage.net1.976.2372.349.367373.130Domaininfo
5yahoo.com2.707.4983.025.174317.676Domaininfo
6wordpress.com1.696.2461.941.639245.393Domaininfo
7markt.de312.879508.875195.996Domaininfo
8apple.com427.187619.440192.253Domaininfo
9amazon.com901.8841.080.874178.990Domaininfo
10yopi.de823.524997.778174.254Domaininfo
11stern.de398.104543.696145.592Domaininfo
12zeit.de403.154530.112126.958Domaininfo
13suchen.de372.214473.450101.236Domaininfo
14faz.net473.143574.302101.159Domaininfo
15meinestadt.de902.981983.84980.868Domaininfo

And while we're at it, here the list with all those domains that lost noticeably due to the Mayday-update:

Losers
#DomainBeforeAfterLosses
1ebay.de3.635.8512.645.512990.339Domaininfo
2preisvergleich.de933.573321.894611.679Domaininfo
3germanblogs.de684.718155.053529.665Domaininfo
4wikio.de944.819454.452490.367Domaininfo
5folkd.com666.508183.144483.364Domaininfo
6oneview.de813.825346.051467.774Domaininfo
7websitewiki.de641.226220.494420.732Domaininfo
8yatego.com1.249.758855.726394.032Domaininfo
9shopping.com1.248.120915.624332.496Domaininfo
10yigg.de320.9923.882317.110Domaininfo
11berlinonline.de586.298269.459316.839Domaininfo
12quoka.de964.070662.460301.610Domaininfo
13hotfrog.de1.258.9691.016.121242.848Domaininfo
14libri.de577.631344.868232.763Domaininfo
15linkarena.com256.48926.748229.741Domaininfo


When I compare the sites on the lists with my “perceived” quality of their content, then it seems that this update was a success for Google. The projects with unique content now show a stronger presence in the longtail, while those with rather thin content, which is also often not unique, are among those domains that lost the most. It is still not clear which dials have been tweaked, though. I could imagine Google putting more emphasis on onpage-factors for the longtail again, while we also cannot rule out that they are using (sidewide) factors like bouncerate and the time visitors stay on the domain.
Johannes Beus - on Wed (08/25/2010) at 12:22 PM

TRG OBS, Irgendwie und SEO, Dmexco & SEOktoberfest

Taking a peek outside the window gives us the unmistakable certainty that summer is about to end, which also means that the frequency of events that are relevant to SEO soars again. Last week saw the start with the TRG Online Business Summit. About 100 invitees discussed the developments and chances on the internet in a relaxed ambience. Personally I was well pleased with the summit and hope for a continuation next year (full disclosure: I'm a partner in TRG).

Stefan is turning 10 and with the help of Bing, he plans on celebrating this birthday in Munich. Going by the name Irgendwie und SEO, there will be a short conference on September 10th, with a large party afterwards. I am looking forward to my time in Munich, seeing how suchmaschinentricks.de was one of the first starting points when it came to SEO. Sadly, there are no tickets left, though the 20 year celebration will be here in not time.

I am not the biggest fan of the OMD Dmexco, which also kept me from being sad, that last year, both this conference and the SEOktoberfest happened simultaneously. This changed this year and since I haven't yet come up with a good excuse, I will probably take a trip to Cologne. At least there will be enough entertainment: on the eve of the conference, there is the Affiliate Regular's-Table by metaapes, after which you can choose between the TRG Online Regular's-Table Cologne or Randolf's OMClub.

One week later I will be back in Munich to conclude the month with my highlight: the third SEOktoberfest. I am looking forward to new faces and to how Marcus plans on topping last years' event – it's the 200th Oktoberfest, after all. On day two, they will once again hold the Friends & Family Dinner, so all who are interested can still bid on a ticket. 100% of the profits from this auction will be donated to a good cause, which means there are no excuses not to bid.
Johannes Beus - on Wed (08/18/2010) at 11:14 AM

The impact of Google Caffeine

It's been a while, since Google announced that they will start using a new search-infrastructure by the name of “Caffeine” as a base for many Google services, in the future. This new infrastructure is now in use and I worked up two evaluations to show the changes this has brought. The first diagram shows the percentage of keywords which have at least one “current result” within the first 10 results. Those are results for which Google also displays how long ago, in minutes, they found or refreshed the site:


You will notice that from December 2008 till December 2009, their quota was pretty steady, which changed majorly at about newyear. By now, about 20 to 35 percent of all searchresults have at least one “current result” on the first SERP. The second diagram shows a similar evaluation, though this time, it looks at the integration of the real-time-search. This is a box, which updates itself constantly with the most recent Twitter-messages and Facebook-statuschanges that pertain to the queried keyword. Seeing how Google only started to use this feature in 2010, we keep to that timeframe:


Even though they started out hesitant at first, they are using this feature more and more now. The waves in this diagram lead us to assume, that they use the troughs are being used to evaluate how well the users are reacting to the feature. It seems that the results are favorable, seeing how the box is being used more often and it seems that we will have to get used to it being there in the future. As the importance of this feature goes up, it's going to be interesting to see how Google will deal with the spam-problem – the amount and quality of ranking-signals is extremely limited with data so fresh.
Johannes Beus - on Tue (08/17/2010) at 22:21 PM

IndexWatch 07/2010

The next productive month is through and it's about time to take a look at the winners and losers of Google's index. As usual, I put together a list of 15 domains for both categories, based on the SISTRIX VisibilityIndex and give some insight into the more interesting positions. We start out with last month's winners:


This month we have a number of domains that managed to rid themselves of a penalty (monstersandcritics.de and spruechetante.de for example), as well as a few domains that managed to garner a lot of visibility in the German SERPs within a short period of time. Cyberport.de for example managed to double their VisibilityIndex score within the past two weeks and afterbuy.de shows a similar jump. If we take a look at the later domain and break it down by hostnames, we can see that, for the past two weeks, Google seems to put more trust into the domain as a whole and therefore ranks all the sites accordingly. We can also see this when we take a look at the ranking-distribution: ever since 07.26. there are notably more keywords present on pages 1 through 3.

Commerzbank's private-customer page profited from being redirected to from the Dresdner-Bank page, Google.at has (once again) won the duplicate-content-fight for the category “translations” against Google.de and Google.ch. Congratulations.


On the losing end, we have a mixed bag of penalties, domain-moves and other problems. Wedding.de has been redirected to a new portal, but was unable to prove to Google that the new domain is as valuable as the old one – about 50% of the former visibility is still missing. While Flyamo.de got away with a warning shot a few weeks ago, they've now managed to catch a full blown penalty: next to no keyword can be found anywhere before position 90.

FHM decided to get rid of their website and only use the Facebook fanpage from now on. This shows in the old domains' visibility – in next to no time, it went from a value between 3 and 4 back to 0. When you take a look at the Facebook fanpage on the other hand, you will notice that – at least from a SEO-point-of-view – this was not such a lucky pick. The visibility is around 1,3.
Johannes Beus - on Fri (08/06/2010) at 11:04 AM

Twitter the second largest searchengine?

For a while now, we've had the same game every month: Twitter announces how many searchqueries they've processed the month before and some quality-journalists put these number in relation to data from Google Yahoo and Bing. Then Twitter gets crowned Google-killer. Something is not right here, is there?

The first problem arises from the fact that the data used comes from entirely different sources. To get the number of searchqueries for the large searchengines, you generally use an evaluation from comScore or a similar, external provider. Those will hopefully use a meaningful panel themselves, which gets you data that can be compared to each other – but not with external numbers.

The second problem is the completely differing definition of “searchquery”: for the searchengines this is rather clear – the user enters in a keyword, finds whatever they were looking for and you get your completed searchquery. Sadly, Twitter counts something completely different. When Spiegel Online integrates a Twitter-stream-box within their left navigation during a World Cup game with German participation, then Twitter will count every Spiegel page-view as a searchquery. The same is true for the numerous API-clients: my Tweetdeck “searches” for “sistrix” every 5 minutes, which makes about 250 searchqueries a day, more than 8.000 a month. Is this a value that we can compare?
Johannes Beus - on Fri (07/09/2010) at 11:02 AM

IndexWatch 06/2010

It is July already, which means it is time to publish the IndexWatch-numbers for last month. As always, we use the changes in the SISTRIX Visibilityindex to present last month's winners and losers in the Google-index. Let's start with the winners:


There are no big surprises on this list. As usual, we have a few domains that escaped a penalty and have a halfway decent ranking, once again. travel3.de as well as celeb.ag are two of those. Pornblog.at is also back in the searchengine's results-pages – here it seems that the “porn-filter” got the best of it and took the majority of keywords out of the ranking for a while, which can be seen nicely in this screenshot.

Otherwise, we have a surprising number or sites with feeble content: obscure Industry- and Company-directories as well as similar offers. I want to take this opportunity to explicitly state that with the above, I do not mean the domain “cylex-review.com”. The last time I made a similar comment, they went ahead and send me a cease-and-dessist letter. The only reason why supportnet.de is on the list is because they had a noticeable but sudden "sagging" of their Visibilityindex-value at last month's deadline – all in all, the domain is on a steady decline since the beginning of the measurements. Sportschau.de is happily trading places with sport.ard.de in the rankings (duplicate-content be praised) but I believe that I have already touched on this subject before.


This month's losers are dominated by domains that fell victim to Google's housecleaning that I already talked about. We can, once again, make out certain parallels concerning the domain's penalties, but it is impossible to make out one definite Reason behind them. We should not assume this to be a one-time occurrence and in the future, we can expect Google to regularly remove those projects from their results that go too far in one way or another.
Johannes Beus - on Thu (07/01/2010) at 08:21 AM

Skoda.de - with full throttle out of the ranking

Even though my Jetta is a source of joy on a near daily basis and the 1.6 liter engine purrs as though it's come fresh off the assembly line, from time to time, I do look at the useful cars from VW's Czech-Brand. It seems that they have finally decided to relaunch their rather antiquated web-appearance. While the design of the new website might be quite pleasing, they made a number of mistakes when it comes to the sites implementation of SEO, which I want to look at more closely in this post, so that others will not repeat them. First, we take a look at the domain's Visibilityindex:


Sichtbarkeitsindex Skoda.de

Since the beginning of 2008 there has not been much of a change, sometimes the rating went up a little and then it went down a little, again. All this changed last week: Skoda lost 60% of its visibility - from nearly 6 points down to 2,3 points. If we take a look at the reasons behind this drop, it soon becomes apparent that the relaunch hit the whole website. Besides the few generic search-queries for which skoda.de was formerly in the Top-10 („gebrauchtwagenbörse“ [second-hand car marketplace] from 9 to 19) it is mostly the traffic-strong Brand-keywords that took a dive. The domain fell from position 1 to 11 for „skoda octavia“. When we take a look at the current, daily results for the keyword, we can see that it is still falling: Position 90. We can expect the Visibilityindex to down dramatically once again when the next update comes around. A good example of how the website-operators did not concern themselves with the requirements that the searchengines put forth, can be seen with the switch from „skoda.de“ to „www.skoda.de“. Additionally, with the relaunch, they also added some hurdles which makes it hard for even the most benevolent crawlers to gain access to the content.

Session-IDs
At first, I could not believe it myself, but the people at Skoda seem to think that they cannot forgo the use of session-IDs when presenting cars. Searchengine-crawlers do not get a different treatment, which means that Google has a number of pages with that parameter in its index:

Session-IDs bei Skoda.de

For searchengines, this yields a number of problems: with every new session-ID that the crawler gets fed, every page on the domain gets a new URL, which leads to duplicate-content problems. The calculations for the internal linkage is also affected by this because new links will turn up all the time while the old ones will have vanished.

Pagetitle, choice of words
When your product is commonly known as “Skoda Octavia”, how would you fashion the title of the page depicting that product? Well, Skoda seems to use a rather unconventional method: the title of every page on the domain starts out with “Skoda Auto Deutschland GmbH” and only then will they use variable titles. This means that once we get down to the product-level, you get a title that does not feature “Skoda Octavia” in that order – not once on the whole domain:

Komische Title-Tags
And as if that would not be enough, it seems that they also have a problem with putting the correct product-name into the actual text on the page. It might be due to some kind of corporate-design-guidelines when they write: “ŠkodaOctavia Limousine” - yes, exactly, without the space. As a result, Google is unable to correctly match the query, which can be nicely seen when you look at the site's cache.

JavaScript, Flash, 404-status-code
While the first two problems were surely the initial reasons behind the dismal Visibilityindex-ratings, there are a number of other problems with the site that make it hard for searchengines to get to the content. The sitemap for example, which usually is a good starting-point for crawlers to discover the contents of a site, is written in JavaScript and therefore invisible to searchengines. Large parts of the website are dominated by a large Flash-area – the text in that area is usually unreadable for searchengines and will therefore not become part of the evaluation. Another problem is the fact that every URL on the domain will return the statuscode “200”, which means “everything is fine”. This becomes especially apparent with the (nonexisting) robots.txt file, but also tempts others to push non-existing pages into the Googleindex ;-)

It is going to be interesting to see how Google will deal with the domain in the coming weeks. Seeing that Google did not start crawling only yesterday and that they also learned that people will make every possible mistake that can be made, the Google-crawler can deal with most of the mistakes rather well and will usually figure out what the website-operator actually wanted to accomplish. I think that this might happen with Skoda – the domain is just too well linked to as that Google will keep ignoring them indefinitely.
Johannes Beus - on Fri (06/25/2010) at 08:18 AM

Google goes after E-commerce

Google continued to expand its activities in the shopping and e-commerce sectors in the past weeks and month. It seems as if they are no longer content with purely being a broker between searchers and online-shops but that they would rather have a larger piece of the pie themselves. For this, they redid the Google Commerce Search for example. Version 2.0 of the internal search especially for online-shops has some new features under its belt and costs only half of what it used to. This should increase circulation and score important data for Google. At Google Products, former Froogle, there are also exciting changes: While the amount of searches with a Google Shopping-integration only went up slightly, as can be seen in this diagram; the type of integration is becoming increasingly different: in the past they contained direct links to the shops, while now, many of the links will be deeplinks to the Google Products detail-page. This is where Google build a full fledged price-comparison-servince over the past few month, complete with product-summaries, scraped user-impressions and the prices from participating shops. This is also where they are now starting to move their universal-search-integrations for nearly all high-traffic keywords to., which can already be marveled at in the USA.

Not nice? If we believe Sameer Samats (productmanager) statements in a current article on CNET, then this will only be the beginning. They managed to build the largest and most complete product-catalog in the world, thanks to the help of more than 100.000 shop-operators, all of which are submitting their current prices and product-details on a daily basis. Though at the moment, this only helps if the users already knows what they want to buy. Right now Google is thinking about a way to make shopping less of a task and more of an experience, to get customers to bind them selves closer to Google. Right now, they already offer an Android app, which lets a user scan a products barcode to get more details. Google is also offering a perimeter search for cell phones, which lets you find shops around you that carry a certain product. For the future, Google is planning the personalization: When it comes to shopping, Google wants to move away from just search and find a way to recommend users new, fitting products. I excited to see how the new Google Products and the lively discussion with data-protection-agencies will play out.
Johannes Beus - on Wed (06/23/2010) at 11:41 AM

Most widely used IP-addresses

There are regularly discussions whether one domain on an IP-address can have negative repercussions for another domains on the same IP-address if it does not play by the rules. For a long time now, I am an advocate of the position that Google is unable to make the IP-address a negative factor in their algorithm because of the large number of shared-hosting-installations that host a large number of domains/hostnames on the same IP-address. Back in 2006 and 2007 I already looked at this and now that I once again have fresh data, I made a top-list with the most widely used IP-addresses, which I want to share:

IP-addresses that host the most domains
#IP-addressOperatorDomains
180.150.6.143T-Online Hosting82.157
274.125.43.121Blogger.com81.686
381.169.145.70Strato60.050
481.169.145.68Strato59.563
581.169.145.75Strato59.541
681.169.145.72Strato59.472
781.169.145.67Strato59.320
881.169.145.73Strato59.281
981.169.145.65Strato59.159
1081.169.145.74Strato59.125
1181.169.145.71Strato59.124
1281.169.145.69Strato58.753
1381.169.145.66Strato58.718
1464.202.189.170GoDaddy.com56.957
15208.73.210.28Oversee.net53.577
16216.8.179.23NameDrive47.947
17205.178.145.65Network Solutions44.028
18188.40.70.45de.vu und Vergleichbares35.647
1968.178.232.99GoDaddy.com27.676
2068.178.232.100GoDaddy.com26.133
2189.31.143.1United Domains21.816
22207.217.125.50EarthLink20.087
2364.95.64.198NameMedia17.729
2485.237.85.69Powerweb17.254
25216.240.187.175Domainsponsor17.129
26209.157.71.50Homestead16.587
27216.240.187.145Domainsponsor16.158
2864.29.151.221Carrierzone.com15.849
29213.186.33.2OVH15.079
3089.31.143.101United Domains15.068
31213.186.33.19OVH14.535
3281.91.170.22Denic14.286
3362.116.163.221InterNetX GmbH13.929
3482.98.86.163Sedo13.578
3568.142.205.137Yahoo13.094
3689.31.143.12United Domains11.622
3782.98.86.171Sedo11.516
38216.39.57.104Yahoo11.426
39216.34.131.135fabulous.com11.179
4081.169.145.87Strato11.130
41216.39.57.107Yahoo11.067
4281.169.145.82Strato10.963
4381.169.145.85Strato10.793
4481.169.145.88Strato10.792
4589.31.143.2United Domains10.773
4681.169.145.83Strato10.767
4781.169.145.84Strato10.751
4881.169.145.81Strato10.727
4981.169.145.91Strato10.718
5081.169.145.86Strato10.671

Funnily, Google's Blogger.com is already on position two thanks to all the blogs that are hosted there, beaten only by T-Online. Next up are GoDaddy and then Strato's mass-servers. These are a great example, since Google can not simply punish each of the 55.000 domains on a server if only a few of those domains do not play by the rules. This is not to say that they will never do a manual examination of these IP-addresses – but even then, I believe that Google has better factors that they use.
Johannes Beus - on Tue (06/22/2010) at 14:38 PM

How is the linkgraph evolving?

It has been more than ten years since Google made links the key ranking-criterion – since then, a lot has happened: PageRank lost some of its significance in the mix of signals, while other factors like “trust” or “link-diversity” probably gained some. You can file pretty much all of these measures under “fine-tuning” – there have not been any fundamental changes in all this time. Getting some fresh wind into the searchmarket would probably be a good thing though and might shift market-shares more towards a fair competition. At the moment, I can make out three possibly directions in which things might progress:

Links stay as main ranking-criterion – things stay the way they are. The Linkgraph still is the primary signal and changes will only happen in how it and other partial aspects are weighted. Google's ability to keep up and even expand their qualitative headway against their competitors in the past few years strengthens this possibility. It does not seem as if there will be an end to this development, with new patents from Mountain View going in the same direction.

Semantics finally make a breakthrough – for years we have announcements of the huge breakthrough being right around the corner, while nothing really happens. Wolfram|Alpha did not herald in a new age, Powerset became part of bing and seems to have gone under. If there was a real innovation here though, I think that a semantic analysis of websites could give us high-quality ranking-signals and at the same time (at least partially) take over the role that links have at the moment.

Socialgraph as a counter-draft – when Facebook let its link-button out into the wild a few weeks ago, it might have caused a stir in other sectors but it does not seem as though the knowledge of a possible impact on the searchenginemarket has not quite reached Germany. Basically, this system – like links – deals with recommendations: Person A recommends URL B. Additionally to that, the Socialgraph has a few advantages: it seems as though thanks to lower “participation-obstacles” you get more data (a blogposting here gets maybe 50 likes but only 2 to 3 links) and the recommendations could be weighted in a way that takes into consideration the relationship between the recommender and the recommended.

Personally, I think we will live with the Linkgraph as primary ranking-criterion for a while longer. The quality of the results has increased over the past years and is still being worked on. I think the Socialgraph is really interesting, but it will be some time before we will see the first uses and results.
Johannes Beus - on Fri (06/18/2010) at 16:44 PM