How long may a linktext be? (part II)
Due to the fact that the test in the last posting did not yield a definitive result, I set up a new test and I am confident that I have figured out the answer. Just as I did with the text a few days ago, I generated a few pages and links to check how long the linktext is allowed to be. Since the previous test has shown that the pure length of the linktext is not a limiting factor, this time I tested the number of words in the linktext which can be seen online over here. For 8 words or 7 spaces respectively, the target-page can still be found through the last linktext:
For 9 words or 8 spaces respectively, it is not found anymore:
You can still find the target-page when searching for the second-to-last linktext (the 8th word) which proves that Google set the cut-off point after 8 words. Incidentally, this is the exact same number that Shaun mentioned in his original blogposting – he just came to the erroneous conclusion that Google is limiting the linktext through the number of characters. I am pretty confident that the two tests have shown that Google is actually using the number of separators.
The consequences that this realization entails are a nice and surprising display of how much in the field of SEO is connected – in this case linktext-length and URL-design. Many users are linking URLs with the URL itself as the linktext, either because they are in a hurry or because the CMS is set up that way. If we are to assume that Google will not only use spaces are separators but also the regular separators, then this might cause Google to ignore the interesting parts of descriptive URLs for the linktext. An example:
I colored the separators in red and, as you can see, in this case the actual “keyword” will not be used as a linktext at all because there are too many separators before it. In short: keep the URL short, put IDs – if necessary at all – at the end of the URL, limit the number of subdirectories to a bare minimum and place the important keywords as far in the beginning of the URL as possible.
[update 22:30] the question of whether I had also tested the idea that this could also have an effect on URLs as linktext came up in the comments. I have to confess that I did not do that. I will make up for it, here is the test: Separators and the treatment thereof.
How long may a linktext be?
A few weeks ago, Marcus at SeoFM reported of a blogposting in which the author assumed that Google will only read out a maximum of 55 characters per link. I thought this to be quite interesting, seeing that – even though the assumption about Google not using overly long link texts seems to be obvious – there are not many informations on this subject which are, at least, partially verified. Sadly, Shaun did not mention the source of this consideration, so I had to make up a test myself.To test for the maximum length of the anchor-text, I created 54 pages where each one receives one link with the linktext-length between 10 and 64 characters. You can take an online look at it here. Seeing that the linktext does not occur on the on the linked-to pages themselves, the deliberation here was that Google will only be able to find the linked page through the linktext if the complete linktext was evaluated. For the 10-character-link it looks like this:
Besides the summary page, which was used to link the pages, the linked-to page was also found. When you click on cache, you get the anticipated info that “These terms only appear in links pointing to this page: sis8f14e45”. I thought that I would just go through the links one by one and, for a certain number, I would notice that Google would not be using the (complete) linktext anymore. Sadly this did not work out at all, even if a link has 64 characters, Google will use the complete linktext:
Because the longest German word (“Rindfleischetikettierungsüberwachungsaufgabenübertragungsgesetz“, at least that is what know-it-all-pedia says) has only 63 characters, a limit that is above these 64 characters should be irrelevant in practice. The next consideration was that link texts are often composed of not only one word and that Google might cut something off for compound link texts. To keep it short, this does not seem to be the case either. The linktext which was composed of two words with 64 characters as well as a space in-between (a grand total of 129 characters) is evaluated without a problem and the cache also holds no surprises.
Now, another consideration is if Google will limit the linktext if there are a certain number of words in it. Sadly I did not think of this when I created the first two tests, but I just now made up for it. Once the results are in, there will certainly be a follow up to this article.
Fear, uncertainty and doubt
Searchengineoptimization has always been a field in which well-founded insights are a scarce commodity. There may be a number of statements that finds a consensus with most people in the field but there are still large tracts which are subject to personal estimates and considerations. This leaves a lot of room for speculation – and I want to write about exactly that point. Many people are accusing Google of intentionally sowing confusion and uncertainties. I rather think that the ones sowing (false) rumors are blogs, forums and companies – be it due to ignorance or because they are interested in certain developments. For a few days now, we have another such case haunting the blogs and the media:NTT Europe, a Europe-wide provider for hosting solutions, commissioned an experiment to evaluate how much local hosting influences Google-rankings. The experiment consisted of 25 searchqueries that were entered by users from the respective countries. They first used the default-search and then the choice “sites from Germany”. Then they resolved the country in which the first five results were hosted. The gentle reader might already suspect this, the result was that, for searches for “sites from Germany”, Google usually found sites that were hosted in Germany. Seeing that this statement is not quite so surprising, even when fitted with a lofty slogan, they had to come up with something else. They bumped into the Hitwise statement that about 14 to 30 percent of Google-users are using the “sites from countryXY” feature from time to time. Combine both and you have the finding that you could loose up to 30 percent of the traffic if the site is not hosted in the respective country. Even though the generalizations and logical-leaps should already have been enough, NTT Europe decided to just turn the whole story around and keep generalizing it for the headlines so that they can now claim: “NTT Europe Online: local hosting enhances Google-rating by 30 percent”.
This statement is now on the table. It is very rare for anyone to take the time to scrutinize the actual procedures as well as the deliberations that went into this statement. This is also the reason why so many blogs are putting this story online without reflecting upon it, which, in turn, gets this nonsense slowly integrated into the “SEO-general-knowledge”. I was a little surprised this morning when I saw that the Handelsblatt had adopting this subject and how they dealt with it. Being truncated and generalized again, under the headline of “server-location decides Google-ranking”, I get to read that “the relevance of websites in Google [...] is determined through the so-called PageRank” and that a local country-domain has a “verifiable, positive effect on the results for a search in Google”. What utter nonsense! Here we have major parts of a press release that get adopted 1:1 by a journalist, inaccuracies as well as untruths which are bolstered as well as added and the whole thing is whisked up into a text that can not live up to the requirements that are posed by a reputable financial newspaper.
Please do not get me wrong: I am pleased with any contribution that gives thought to the functioning and mechanics of Google. Tests and analyses are the foundations of solid SEO-work. But for an industry which is dominated by so many uncertainties, assertions, like those posed by NTT Europe, are extremely counterproductive. I wish that bloggers and forum-user would scrutinize such statements a little more acutely – while I actually expect this from journalists.
How is Google reacting to the crisis?
There is no need for a crystal ball to predict that, for a majority of companies, the next quarters will not hold up to the expectations that were set in the beginning of the year. There is also the consensus that the web-industry will be affected, too. The way in which Google – who is, after all, flush with money and on top of the industry – is adjusting to this situation is something that I think is very interesting.On the one hand, they are trying to increase revenues on many different fronts: they are allowing advertisements for beer and, by now, also for hight-proof alcohol, Google is allowing gambling AdWords in England, we are also seeing advertisements pop up on Google Maps, yesterday AdSense opened up to domainparking and it seems that they are even trying to hasten the commercialization of YouTube. On the other hand, Google is cutting costs. It was quite a surprise when Google announced that they are letting temporary workers go (where it is interesting to note that, of the 30.000 employees, 1/3 are temporary workers).
It is exciting to see if this move has any influence on the innovativeness and the development at Google. This is an area where Google has an enormous advantage over competitors of about the same company-size and I can imagine that if Google shows any weakness, many interesting chances will open up.
Linkbuilding
Yes I know, the title for this posting is rather generic, but I thought a bit about the problems and risks involved in the current linkbuilding-techniques and- Lack of market transparency – With its fight against links that are not set solely as recommendations, Google caused the complete marketplace to proceed in a direction which is opposite to the needed transparency. Private link-networks and sources are more important than ever while public marketplaces like Teliad or Text-Link-Ads are becoming less important.
- Uninformed customers – Thanks to the present SEO-hype, there are many customers who are buying SEO because “that is what you do now”. I am convinced that, especially in the area of SEO, the customer should have a solid basic knowledge of SEO at their disposal to be able to assess the performance of the SEO-agencies. I will not only accuse the customer, as I see it, the SEO-agency also has a duty to inform their clients.
- Short supply, high demand – I does not come as a surprise that the availability of “good” links is strongly limited and they are getting increasingly rare due to Google's actions. At the same time the demand for links is growing rapidly.
- Conflicting goals – usually the SEO-agency will be the seller of linkbuilding-measures and, at the same time, consultant on the reasonability thereof. This is not a good combination – especially not if the client is uninformed but still trusting to be correctly advised. Here, it seems natural for the SEO-agency to advise with their own profit-optimization in mind instead of seeking the best solution for their clients. This is particularly true when – having happened more frequently in the meantime – the Onpage-optimization and consulting-services are rendered for free, if the customer purchases the linkbuilding through the same agency.
- Short-term actions – Solid SEO-measures have more of a 3 year horizon than one of 3 month. This does not stop many service-providers from capitulating to short-term target-agreements, which they will try their hardest to uphold – regardless of the methods used. This leads to the application of linkbuilding methods that anyone who is able to, at least, think partially clearly would never even consider and which will massively harm the domain sometime down the road.
Dates, data and statistics
A minute ago I took a look at the SEO-dates for the upcoming year. It all starts on the 7th and 8th of March with the “SEO Campixx” in Berlin. The Seonaut has created this as a two day SEO-Barcamp and I am curious of how much of the Barcamp-concept will be realized in the end – as far as I can tell, the program already looks quite planned. The “must go” event per se, the SEMSEO and the Abakus Pubconference 2009 are on May 15th of next year. They are, as always, in the quaint Hanover and I am sure that Alan will manage to make this year great again. I am not quite sure if I will attend the two “default-conferences” (the SMX in Munich and the SES in Hamburg). Content wise, both are not that interesting to me seeing that they are more about networking and in such cases I will decide on short notice, depending on the participants. The September will prove to be exciting with the new dmex in Cologne going up against the OMD in Düsseldorf. Both of them are within taxi distance, so lets see which one hosts the better party.Randolf has found an interesting download at Alexa (yes, those with the ineffable toolbar): the Top-million-domains by Alexa-rank as CSV-file. Up until now, Alexa has been selling these files for a tidy bit of money, but it seems they are starting to be more generous with the competitor to Google (Trends). As far as this is concerned, it is interesting to not that Alexa is not just counting Domains anymore, but is trying to track sensible smaller units. While this still makes sense for the subdomains of Webbloghosters like Wordpress, they they are overshooting their mark with individual capture of Feedburner-feeds (feeds.example.org/~r/) – in this case Alexa seems to divide along the tilde as the “classic” signal for the home-directory of Unix-systems.
Today, Seomaxx published the results of a survey that the visitors of their booth at this year's OMD filled out. 279 visitors filled out the survey after all, here briefly the most important numbers: 69% are not yet practicing SEO, 31 % already are. They are spending an average of 8147 Euro a month for SEO, which is a reputable amount if you ask me. If we take this and put it in relations to many SEM-budgets, then I think that there is still some room for improvement. Those of you who want to get their hands on the complete survey can write Michael an e-mail.
IndexWatch 11/2008
Now that the lousily cold and, even for Bonn, snowy November is over, it is time for latest IndexWatch-numbers. As a reminder: each month, with the aid of the numbers from the SISTRIX Toolbox, I compile a list with winners and losers in the Googleindex. Older postings can be found through the tag. First the winners:winners
| # | Domain | Veränderung | |
|---|---|---|---|
| 1 | hotelkatalog24.de | +1.561,0% |  |
| 2 | imdb.de | +1.165,5% | |
| 3 | edelight.de | +132,4% | |
| 4 | linkarena.com | +87,6% | |
| 5 | dawanda.com | +81,2% | |
| 6 | buchfreund.de | +75,8% | |
| 7 | verivox.de | +71,0% | |
| 8 | softguide.de | +69,1% | |
| 9 | acont.de | +64,5% | |
| 10 | shortview.de | +63,6% | |
| 11 | drweb.de | +60,5% | |
| 12 | favoriten.de | +58,9% | |
| 13 | medikamente-preiswert-bestellen.de | +54,4% | |
| 14 | abconline.de | +53,5% | |
| 15 | auvito.de | +51,7% | |
In November, after a prolonged phase of weakness, hotelkatalog24.de has returned to the level that the domain had until the end of June of this year. The prolonged sitting out seems to have paid off in this case. Imdb.de is noticeably profiting from the fact that IMDB(which
losers
| # | Domain | Veränderung | |
|---|---|---|---|
| 1 | automarkt-gebrauchtwagen.com | -98,1% | |
| 2 | swoodoo.de | -95,1% | |
| 3 | tripple.net | -94,2% | |
| 4 | openphrases.com | -92,5% | |
| 5 | chirurgie-portal.de | -83,0% | |
| 6 | tipp24.de | -80,7% | |
| 7 | cosmoty.de | -60,0% | |
| 8 | evendi.de | -54,4% | |
| 9 | crazylifeblog.de | -50,0% | |
| 10 | gmx.de | -44,3% | |
| 11 | top-dsl.com | -44,0% | |
| 12 | branchen-info.net | -42,1% | |
| 13 | koeln.de | -41,6% | |
| 14 | freepatentsonline.com | -35,4% | |
| 15 | linx.de | -33,1% | |
On the losing side we have quite a few “classic” SEO-projects this month. I would like to use automarkt-gebrauchtwagen.com (what a domain) as a model and retrace their crash. Up until the week of 11.17, the domain ranked for a number of thematically interesting keywords in the top-10: Gebrauchtwagen (pre-owned car), Automarkt (car marketplace), Autokauf (car purchase) and so on, which should have led them to pick up quite a bit of interesting (=lucrative) traffic. During the week of 11.17 to 24.11 this was all over with. In a blog posting a few days ago I have shown how the arrangement of keywordpositions can be a decent indicator for a Problem with Google and this domain shows a similar picture. While before, about 30% of all keywords for this domain were on the first SERP, after this date, nearly none of them are there anymore. It gets interesting if we ask for the reason behind this: since this site is not crazily extensive and since, on first sight, I was unable to make out any serious OnPage-mistakes which could justify such a penalty, I took a look at the backlinks and that is when I came upon a multitude of links that were placed as counter on foreign sites (example 1, example 2, example 3). In their Webmaster-blog, Google already warned about exactly this constellation a few month ago and now it seems as if they started to follow up on that.
On PageRank-changes
Seeing how the stock of topics is rather meager at the moment and how I thought that rather than saying something spiteful on the topic of linkbuilding via commentspam for large brand-customers in the insurance- and shoppicsector or on the state in which an ominous SEO-scene is in general, the evergreen of SEO-subjects would be more suitable: PageRank, or rather the changes therein. At the moment there are three reasons why the PageRank, for a URL, which is shown through the Google-toolbar, changes:PageRank-update
Google exports the internally used PageRank to the datacenters in, by now, quite regular intervals and is therefore responsible for the notorious PageRank-updates. The last time that this happened was on 09.27, before that on 07.27 and 04.29 – which means that the odds are in favor of us seeing another update before new years. In general there will be no change in the ranking after a PageRank-update. The “new” numbers were known to Google for quite a time and were already integrated in the ranking. One point that is quite interesting is the reliable detection of PageRank-updates. For this you take a clean site that is a little older that is regularly publishing new articles. Blogs like this one here are especially suitable for this. Then you go back through the articles until you find the last one that still had PageRank during the update. As luck would have it, for me that is the posting about Bernd Sonnensegel. The content which comes right behind it chronologically (this one), is currently shown with a gray bar in the toolbar, which means that it is without a PR-value (for these the German blog postings are used as the translated ones are both without a PR-value at the moment). If we get a PageRank-update, this value will change to at least a PR of 0.
PageRank-adjustment
Besides these large updates there are a number of smaller adjustments of PageRank for single sites and domains. This can happen if Google detects a 301-redirect, if a domain is unreachable for a longer period of time or if they detect two URLs to be identical in content (e.g. www.domain.tld/ and www.domain.tld/index.html). As we can see in the graph on the right, this does not happen that seldom. The graph shows us how many percent of the domains that are monitored by the SISTRIX toolbox suffer changes in the startpage-PageRank. If we compare these changes to a “real” PageRank-update (diagram for 90 days) we will certainly make out a much lower rate of change, but a change all the same. As a result you have forum posts that are shouting “update” to soon, even though there are no new value, just an adjustment of already existing ones.PageRank-depreciation
The third reason is the detection of linkpurchases, linktrade or other demeanor related to linking which does not suit Google. For about a year now, Google is at work to deprive such sites of part of their displayed PageRank. At the moment the “penalized” site does not suffer any consequences and even the realization itself shows some interesting loopholes (A vs B), which leads us to conclude that Google themselves are not taking this too seriously. Right now we can assume that these depreciations are done purely manual, however I was able to observe some interesting values within the last few weeks which led me to assume that Google is also trying out other approaches.
Lycos is closing its doors
With Lycos, we have a former Internet-heavyweight that is stepping down for good. While being the biggest European Internet-portal during the last years of the former millennium and having had success with both the searchengines Fireball and Hotbot.de, it seems that, in recent years, Lycos' lacking adaptability proved to be its own undoing. The restructuring-measures of the last few years did not help and it seems that there was no one who was willing to buy up Lycos as a whole. Now they are trying to sell off their lucrative stakes (United-Domains, Pangora and Jubii) and close down the rest (Portal, including lycos.de and webhosting). It is sad to see this happen, but besides Lycos IQ there was no new product-development that caught my attention in the last few years that could pay for their 700 employees.Clickrequest
Despite a cozy 40° fever, which has sadly also kept me from flying to Munich last Friday, I want to, briefly, point to the first Poker-elimination-tournament within the scope of the Webmaster on the Roof listener-gathering, which will be held in the end of January. The Guestlist is already full, so whoever still wishes to attend can only get in through the poker qualification. So sign up and do not fall for the Pokersucht-addiction ...Additionally, I would be happy if everyone helped me get a license from “Hell Engines”, which can be won at
