How does it work?
With the new canonical tag you indicate that there might be more than one URL within your website presenting this specific information and if search engines want to choose between those when presenting it to visitors, you would prefer the given URL. I for instance use wordpress to rewrite to nice URLs but this post is available under http://www.vdgraaf.info/?p=120 as well.

If I would place <link rel="canonical" href="http://www.vdgraaf.info/canonical-tag-creates-new-legal-cloaking-possibilities.html" /> in the head of this page, Google would only show that one to visitors. But just like a 301-redirect Google would attribute all linkjuice to the remaining page.

The canonical tag is intended for sites that cannot do a proper 301 redirect because of several reasons. For instance if there still are slight differences between the pages under different URLs, but too little for Google to see them as unique. The same article might be available under different categories, an affilliate ID in the URL just changes some form values or you use almost the same text for the French speaking population of Canada as you do for people in France. I can think of many more, but all these legitimate reasons allow for slight changes to the page served. And it is logical that search engines now allow you to indicate which version to choose.

One page, different audiences
Your website is created for several audiences and in the ideal situation you could change nuances depending on the visitor type. The most important audiences we want to differentiate right now are “linkers”, “navigators” and “searchers”.

• People linking to your website
  A clean webpage without much commercial intent is far more likely to recieve inbound links than a page with a lot of branding and call-to-actions. A page that shows your good side (for instance with a reference to your altruism to good causes, or a reference to the specific link partner) will give you a far higher success rate on your link building effort. Use a URL like http://www.vdgraaf.info/i-am-a-good-boy.html for them.
• People navigating from within your website
  People that navigate through your website follow a certain path and you can offer more specific content depending on where they came from. http://www.vdgraaf.info/so-you-clicked-the-banner.html would be good for this audience.
• People from search engines
  A good landing page for search engines gives the answer to what you searched for and shows a clear call-to-action to the most logical next step. Besides changing the appearance of the page that you would like search engines to serve, you could also use slightly different code to be more search engine friendly. You should make this version the canonical one by adding <link rel="canonical" href="http://www.vdgraaf.info/this-one-is-for-searchers.html" /> to the head section of all your versions.

Does the tag give real cloaking abilities
It is logical that Google will or already has implemented a check on simularity between different versions. Because the tag is onpage and not in an external file like the robots.txt, the page has to be spidered first. It is a small extra step to do some text calculation as well. The canonical instruction will probably be followed when the page would normally trigger duplicate content filter. It will however assign all accumulated links to that one remaining occurrence.

So use this tag with care, but with the same content you can still create many different versions. Have fun!

You can now view keyword volume estimates in the Adwords keyword tool!

So far Google has never been “close to” open about search volumes. Google trends only shows charts without numbers and it has a threshold of about 1000 searches per day. Previously the Adwords keyword tool showed only monthly graphs without any reference field to numbers. But at last they changed their ways and show “rough estimates” with real numbers.

In European countries where Google has a marketshare of more than 90% this is close to a revolution. Yahoo data, Trellian, HitWise, Wordtracker and Alexa are all useless because of the limited coverage in those countries. Hopefully Google will improve even more by showing “real numbers” in stead of the current “rough estimates”. But even the current improvement is a welcome change.

Many webmasters have contacted me recently with the same problem. It’s an old problem that is still very important in all search engines. Their websites weren’t indexed entirely or perticular pages couldn’t even rank for the most unique text phrases although these pages were indexed. In this post I will give some pointers on effortlessly making pages more unique.

Large database driven websites all have the same problem. Sometimes search engines do not index them fully. But when they do, many pages still don’t rank in the search results. This is caused by the uniqueness of the content and seeming importance of those pages. Let’s take an average large database driven website with for instance jobs. These sites contain large amounts of job descriptions that are all formatted in the same way. While the content is somewhat unique, the shear amount causes part of that website to be seen as duplicate.

So how does this duplicate content filtering in for instance Google work?
Duplicate content filtering isn’t a black or white issue. It has multiple shades of grey, that in the worst case penalize an entire site and in the best case just effects the ranking of a page slightly. Because all forms of duplication across pages can effect your ranking, it is very important to know how to avoid duplicate content. To ensure perceived search result quality, removing duplicates is high on the agenda of every search engine.

Real focus on a search term is best given by dedicating an entire page to those search terms. This means creating and filling pages can become a huge task. Computer generated or scraped text is a very easy way to create pages, but that is where duplicate content filters often kick in. When you want to rank for combinations between “jobs in” and every city you can think of (for instance “jobs in amsterdam”), you probably generate many copies of the same page and replace the city spot wherever you can. And that is exactly what search engines want to combat.

Duplicate area’s
Many search engines see a page as part of a website and they can distinguish between the header, footer, menu, content block, etc. In fixed blocks duplication is very common, because the header and main menu are usually the same across an entire site. In the content block duplication is less common, so any duplication there is something search engines look at more closely. While duplicate area’s on the entire page should be limited, the percentage of duplicate text in the content block is extremely important.

The more inportant the page, the more duplication is condoned
If your homepage and a page just below it are near duplicates from each other, they can still rank on the part that makes them unique (even on more competitive terms). When the near duplicates are located further down the navigation and they recieve little linkjuice, the chances of them not ranking or even beïng omitted from the index are ever increasing. Linkjuice transfer is very important and optimizing it can fix many duplication issues.

The illustration above shows 2 navigational structures from the homepage. When the homepage gets an extra link on it, pages further down recieve less linkjuice. Less linkjuice means a higher chance of getting caught by duplicate content filters. Put pages higher in your navigation or acquire external links directly to them when you want to make sure they rank in spite of duplication.

Unique mashups
The “jobs in …” example will be easily detected if the city is the only inserted text. So how can you make such a thing work without having to write loads of text? You create unique mashups!

A mashup is a collection of different types of collected content. When you write small pieces of unique text per page and collect all other content in small pieces from many different sources, search engines will love your pages!

In the “jobs in …” example: Write a fifty word intro about “jobs” per city you want to focus on. Add a list of about 10 job descriptions per city from your database. Scrape a piece of city information from a cityguide. Scrape extra pieces of additional information from other sources and finally randomize the order of those content blocks. Try to collect a total of about 300 words. Search engines are smart enough to detect this technique, but the people who use it, have been ranking for ages. The linkjuice to those pages, the amount of used sources and amount of unique text you write determine if you rank on all cities.

Keep the good content above the fold
Unique text is very labour intensive and quality text cannot be automated. But where do you need quality text? Just get your visitor to click a button before they start reading the entire page content and they won’t notice the low quality Focus good usability and text quality on the top part of your page. People rarely scroll and read in detail if the function of the page is already clear and the navigation options are very obvious.

Lazy people can still score with automation, but I prefer using cheap copywriters!

How important is personalized search?
Google (and other search engines) want to give the most relevant search results. Knowing who is requesting the information with what intent is extremely important for this search relevance. If I search for “apache”, I might be looking for the Linux webserver; If my neighbor looks for it, he probably wants the helicopter gunship. So how does a search engine discover: a. the meaning of a word and b. the intention of the searcher?

The meaning of a word
Google doesn’t use a dictionary to find related words (only in a very small way), it uses other occurrances of the search term on other websites to find them. For normal ranking or when a single word can have several meanings (hononyms), these related words become extremely important for the search relevance.

If a large group of apache websites also mention something about webservers, PHP and Linux. Apache must be related to those terms and those websites must be about the same meaning of that word. If another group of sites mention nothing about webservers, but keep writing about geronimo, tipis, crazy horse and tribes, they are probably refering to indians.

Being related
The “occurrances of the search term on other websites” part is becoming extremely important in personalized search. Optimizing your relevance has to do with mentioning the same related words as other websites within your niche. Links to and from your website should contain those same related words. This relates you to your niche and that meaning of a hononym you might want to rank for.

Personalized search testing
After using different identities with a separate search history and behavior, the search results for several hononyms gave a very clear picture of how small changes in your behaviour can have a profound effect on the search results. And the same tests over time, gave us the impression that Google is increasing the importance of personalyzed search. Adding &pws=0 to the search result URL shows you the difference between your personalized results and normal ones.

Besides hononyms we’ve also seen very distinct search results for subniches of a topic. While only using a more generic term for a niche, websites that used related words that coincided with the subniche preference of the searcher where given more weight. So every search is influenced by personalized search in some way.

Intention of the searcher
So how does a search engine find out the user preference for a specific subniche or hononym of a search term? Mainly by looking at your recent searches and things you’ve clicked!

Let’s say you’ve searched for “apache” and after seeing pages full of server results, you re-specified your search to “AH 64 apache” and clicked a helicopter link. The next time you search for “comanche”, search results about the RAH 66 comanche helicopter are given extra weight. In this case some results about the comanche tribe still outrank the helicopter, but suddenly more helicopter links enter the top 100.

SEO for relevance
To become more relevant, you need to find out what you want to be relevant for. Define what audience you want to target and what they define as relevant. If they use terms like AH 64 and technical terms for weaponry, they might be a different audience than the ones using “apache squadron iraq” although they both mean the same helicopter.

Broader ranking is still possible, you will just need more superiority over the closer niche matches. Define the niche and then define the search terms. The exacter the niche, the better your focus and the easier it is to rank (link relevance).

Find out what the most visited (high Google ranking) websites in that niche are and look at the related words they use. Use the same related keywords in both your content, anchor texts to your websites and link to pages in that same niche.

SEO for personalization
SEO shouldn’t be about large amounts of traffic, but about converting visitors to a predefined goal (like “more sales”). Personalization should help you achieve that goal more easily, but you might want to rule more than just your niche. Any traffic is valuable as long as only one visitor converts. Getting people earlier in their decision process, before they know exactly what they are looking for, is becoming even more important with personalized search.

The websites found early in a search process define what more specific search queries someone will be using and they define what Google sees as personal related keywords that need to be given more weight. Therefore top rankers for more generic search terms define the keywords you need to incoorporate. The ones that are the most likely to be “clicked” are more important. The “Searches related to:” tab that is shown at the bottom of some search results helps you to find the terms that are mainly used for further specification.

Conclusion
So far personalized results mean: You have to already rank for a search term the old way, but extra weight is given to specific niche keyword relevance. This niche is mainly defined by search behavior, but the future it may include more personal information to be used.

For SEO’s this means they will have to use better subtopic specifications based on possible preferences and search history. Making a separate page for every subtopic, but mainly focussing on the same search term, might be a good solution if you aim for large amounts of traffic. For spammers it means they should use smarter scraping and text generation software. Personalization is here to stay!

Search engines like Google collect so much data, that there is bound to be something to link multiple websites together. The only question is; with how much certainty can Google then say: “This is the work of the same spammy optimizer”. My tip if you plan to break the quality guidelines: Alter your SEO tactic and everything that could be fingerprinted to you with every website you optimize/build. Here are the most common mistakes of leaving fingerprints for Google to find.

The obvious: leaving your identity (identity prints)
With the use of centralized logins for all Google services and logging of everything you do, Google knows much more about you than you might think. When you use the Google toolbar or login to multiple accounts from the same computer you are bound to slipup and connect your spam activities together unintentionally. When you stay under the radar with all of them, this is not a problem. But when one is caught, the other ones will be looked at more closely. To hide your identity, you need to use multiple identities.

Your scripting style (coding prints)
The most common fingerprint you can leave is in your HTML, Javascript or Stylesheet scripting style. If all your websites have a specific buildup that is somewhat unique for just you, that is a certain print to link them together. Using the same javascripts or HTML pieces in all of them might not showup in normal Google results, but Google’s spamcop tools will certainly find them. Even if you have found a scripting style that works best for you (or the search results), alter it as much as possible. Coding prints can be both in your spam tactic as your normal coding.

Linkerprints
I hope my new definition will become standard for these types of fingerprints. Linkerprints are the common signs in your link structure that tie your websites together. A linkerprint can be used as print when the link sources or targets are common within your websites and more unique compared to other websites.

For example:
Somebody selling linkbuilding services that uses the same link sources all the time is bound to raise some red flags. When this red flag is confirmed, all websites with the same inbound link structure will probably loose much of the value of those links.

Another example:
When you directly or indirectly link your websites together, these fingerprints combined with questionable identity prints are enough to indicate a common ownership and one ban can spread to all your properties. So prevent linking to another property from the first and prevent using directories you moderate to boost all your property. Never link to all your property in one breath!

From prints to penalties
Identity prints, coding prints and linkerprints are used to categorize online fingerprints a search engine can use to find common denominators throughout the websites you optimize/build/own/etcetera. Because search engines need a degree of certainty before they give heavy penalties, they need to have unique fingerprints from more than one category. If they search hard and long enough, they will probably find them. But when finding the prints takes too much time and resources, they will probably give up and leave you penaltyless.

Linking sites together doesn’t always lead to propertywide penalties, but if one is penalized they will take a better look at your other properties to find other spam.

To sum it all up:

1. Leave as little prints as possible (in every category) and make it harder to find them.
2. Make your prints too common to point just to you.
3. Stay under the radar!
4. Alter spam tactics continuously and don’t use common denominators within them.

And of course my main tip is: Don’t spam, just make the best website ever build!

First let me tell you how much impact universal search will have. Google shows these extra results at the top, in the middle and/or at the end of search results. A search for “darth vader” shows it all. Images on top, a movieclip in the middle and related search and news results at the bottom.

These results really stand out and a visitor will more likely click one of the special results. Because they stand out so much, they hugely impact how people perceive the results. If this is something positive for Google usage remains to be seen, but it sure is something positive for us search result influencers. It is like the wild west has opened up again with new frontiers.

Reputation management
Before I tell how you can influence search verticals like image search, I would like to show an example of reputation management issues that will arise.

These are the results for the Dutch prime minister “balkenende”.

Because he looks somewhat geeky like “Harry Potter”, people created images showing him as Harry Potter. His reputation can only benefit from it, but in most cases having the wrong pictures on top of your name or company name search is something bad. I just got hired to boost pictures of a drunk CEO when searching for a company name. Although this isn’t very ethical, I wanted to try it (and I don’t have morals ).

Optimizing image search
No, search engines don’t look at the image itself! Most search engines have little image recognition skills (But look at this example to see what Google can do: image type: face). It would just take too much processing power. Duplicate images, x-rated and other unwanted pictures are filtered with some image recognition, but nothing very advanced.

So what do they look at?

1. Most search engines, including Google, prefer images that are used within the content of the website. Otherwise they would index your entire navigation and design.
   Some believe images after one of the first paragraphs are the most likely ones to be included in image search.
   Pages with texts containing just a few images are more likely to have high ranking images than entire galleries.
   Update: Someone just told me it matters if the image is also a link. He says it is better to have an unlinked image. I haven’t tested this, but it would make sense.
2. The used alt text and text in the immediate surrounding of the image tell the most about its content. Just be concise and descriptive and have it contain little more than the phrase you want to rank for. The title tag of the image seems to have some effect.
   The image name or URL should also contain the search term. Be sure to separate words with for instance a dash, slash, hyphen or dot. Relying on stemming isn’t always good.
3. If the image is contained by a page (and site) that ranks for the search term and also has many related links (with the correct anchor text) to it, that boosts the image ranking very much. Because not many images are optimized intentionally, a ranking page with an image in the content (that might not even contain a correct image) will be the most likely winner for most searches. To even appear in image search the image needs to be on a site with enough links to it in general.
4. An image that is requested from multiple websites (the url on the original source is used to show the image) with the right alt text etcetera has even more chance of ranking.
   I was also told that a copied image (same filename and image properties (bits, resolution, colors, size, etc.)) located on another website, helps the ranking of the original source, but I haven’t seen any proof yet.
5. An image situated on a strong website (for instance a high PR page) is more likely to get indexed. To see all listed images from a single domain, use the site:domain.com command in image search. I hope my site is strong enough to get some images listed, but otherwise I’ll try to boost them via stronger sites.

Example 1:
You should try to get an image of yourself rank for your own name. I haven’t got round to it before, but here it is: In the near future this ugly image should score for the phrase Peter van der Graaf.

Peter van der Graaf

I get many reactions because of my ugly avatar with the green background. I have to say all of this is intentional. It stands out on for instance MyBlogLog and people visit my site to see what weirdo has visited their site.

Example 2:
If this was the first paragraph of a page, it would contain some bla bla on the topic and it should contain the search phrase I wanted the image to rank for. In this case the image below needs to rank for “balkenende search results”. Bla bla as you might know premier balkenende is wel known for his raar kapsel. Jan Peter Balkenende looks a bit like harry potter. Look at the image below for the balkenende search results.

Balkenende search results

This would be the next paragraph with more related bla, but it doesn’t really matter what I type. As long as it is enough text to fill another paragraph. So here I go: Search results are bla di bla and prime minster or Wouter Bos is something related as well. I will type some more to fill some text, but now I am finished.

Check the image search results for “balkenende search results” in a few weeks and this image is sure to score. It takes a lot more time for the image to get indexed, than it does for a normal page.

Matt Cutts (Google’s main spamcop) is back from SES New York and has spoken to too many SEOs. Even with a normal non-spamming website you will need to hire an SEO to check if you’re not possibly breaking any of the “webmaster guidelines”, otherwise Google might ban you from their index.

1. Below in this article: Hidden links Matt says: “As long as we’re talking about links, this seems like a pretty good opportunity to talk about a simple litmus test for paid links and how to tell if a paid link violates search engines’ quality guidelines.”
2. It seems like Matt believes in devaluating the link in stead of the linker in this post.
3. But he also wants us to know how to report paid links on your competitor’s website.

Do you have ads on your website?
Almost 50% of all websites run ads on them and far less than 2% of all webmasters have ever read the Google quality guidelines that so far mention nothing about paid links being bad. But what is the difference between an ad and a paid link? Are all ads bad because you do not nofollow them?

I hope Matt’s article was misinterpreted by many people, because it seems many people disagree in the comments. His wording wasn’t chosen too well and he makes it look like Google will want you to give more information on your intentions. And therefore make part of your code for search engines and not for users. If you don’t, you can face a penalty! Again, I hope they’ve misinterpreted his words, but that is what he’s saying.

Has Google lost the war on spam?
When there is an ever increasing amount of people gaming your algorithm, you are bound to lose from some of them. Google can detect many paid links just by their characteristics. Link buying and exchange programs can be discounted in their calculation and notorious link spammers can be removed from the index. But Google can’t detect algorithmicly when you “in your own way” link to a certain website and get payed for it. And because those payed links have no common characteristic the algorithm can detect them by, there is no way to discount them. This is why Google wants you to tell them if you are getting payed for a link. And if they find a common characteristic in the future and you haven’t told them you got paid (nofollowed the link), they are bound to reduce the trustfactor of your website in their calculations. So they haven’t lost yet, but using threats to get what they want isn’t a sign of strength.

Websites for users or search engines?
Search engines are increasingly telling us: “We have tried, but without your help we cannot order the world’s information the correct way.” “We need the metatags, rel=nofollow, legal cloaking and other information not intended for your normal readers, so please use them but not for evil purposes.”
With Matt’s article Google has taken the next step and he shows this is probably just the beginning. From now on every webmaster needs to know what a search engine can or cannot read and help the engines rank their website for the right content. Is this something good? For SEOs it is

So when is a paid link bad?
If you read Matt’s article and his comments carefully, you will find out that non-obvious ads and paid links, that the algorithm hasn’t detected automatically, are the ones you should mark. Nofollowing every other link that doesn’t help your ranking can even be good for your ranking, but that might raise some red flags.

In a total strategy I will also continue to buy links as long as there is no characteristic that signals “I’m payed”. When there is such a signal, Google should devalue the link and not penalize the linker. Buying relevant links is good for both your visitors and search engines. How else would they know my site is the most important resource on …

Two years ago
If the last time you read SEO blogs or Google press releases was one or two years ago you can still know enough to get good rankings. Not that much has changed in the algorithms for normal websites. The most important changes have been made in spam detection, so optimized whitehat sites shouldn’t be effected that much. But is this entirely the case?

The last two years search engines got better at crawling websites. So if you still obide by the strict crawlability rules of two years ago your website will be perfectly indexed. If you used normal text in the website with extra focus (with repeats, in urls, in titles, in headers and other important places) on the desired search term, not that much has changed either. Placing extra focus on your most important pages with a good internal navigation was as true back then as it is now. Two years ago search engines had already placed more importance on relevant links compared to irrelevant links, but they just got better at determining relevance. So what did change?

• Two years ago search engines had just begun to include time as a factor in ranking and that factor has gotten more important. Many new authority factors make use of time in their calculation. If you’re still used to old SEO, you will need to do alot of reading to grasp how this influences everything.
• Providing robots with extra information how to index and show your website has changed much since then, even during the last couple of months. Google webmaster central, <meta name=”robots” value=”noodp” />, 301 redirects and rel=”nofollow” are just a few changes that improve your communication with the search engines. How would you know this when you have been out of the SEO loop for a while?
• Search engines have gotten better and better at tweaking the search results for specific searchers. Country, language, device accessibility and content type targeting have all dramatically improved during the last two years. Most search engines are now even including personal history and other personal characteristics in their ranking alorithm.
  How do I rank in the UK in stead of the US? How do I rank in Google maps or mobile? How does someones search history effect my ranking? These are all plausible questions for someone that doesn’t read much SEO news.
• Many new search types have been launched during the last two years. Video, blog search, articles, maps, mobile, code, news, desktop and much more new types of search results have arizen since then. Even a seasoned SEO doesn’t know how to optimize for all of them, but they at least know they exist and they remember where they can find more information when they need it. And that is the most important thing you need to keep up to date.

There are probably many more important changes that I neglected to mention, but for someone that has been out of the loop for some time the specifics aren’t that important. But where do you get the information when you need it?! Someone that was an expert two years ago, may be a layman on all the new aspects. Who do you trust and on what topic? The search landscape constantly changes.

The most important thing you need to keep up-to-date is your network of experts and sources. Without it you have no way to really keep up-to-date SEO knowledge.

Google Adwords quality scores
First let me tell you there are two distinct quality scores in use with Adwords. One influences your minimum bidprice and one influences your actual cost per click and ranking.

Minimum bid price
The minimum bid price is the cost per click you have to pay to have your ad appear. Even if there are no competitors, your keyword will be placed on hold if you bid less.

Minimum bid prices are mainly based on:

• The historic relevance of your (entire) site to a certain topic.
• The relevance of the used landing page to the used keyword.
• And the popularity of the keyword.

As you can see you can’t just bid on any keyword with any website and any page. You can, but you pay much more. Bidding on a new unrelated hype just to get large amounts of traffic is therefore much harder to do.

The relevance of both website and landing page isn’t only based on the text, but also on relevant links to them. A page that scores or could score organicly is therefore the best landing page and SEA (search engine advertising) needs good SEO (search engine optimization) now, more then ever.

Ranking quality score
A high ranking quality score gives you a higher ranking or lower actual cost-per-click than a competitor that has a lower quality score.

The ranking quality score is more based on coherence between keyword, ad text and landing page, than actual organic ranking. Both ad text and landing page need to include the keyword and related words. Ad text and landing page need to have simular content so the algorithm can detect if the promise you make in the ad is kept on the landing page and is not some trick to boost the click through rate.

Ranking quality score is mainly based on:

• Historic click-through-rate on the used keyword matching option.
• Historic click-through-rate on the used ad (title, description and landing page).
• Historic click-through-rate on your account, campaign and ad-group.
• Used matching options.
• Bounce rate (direct returns to Google after clicking the ad).
• Ad relevance to keyword.
• Landing page relevance to keyword.
• Coherence between ad and landing page.
• Not as important, but will be in the near future: Organic ranking of the landing page.

It is not as much that you need a good SEO to influence these factors, but you need a good copywriter with expertise in search. Your copywriter should know how to take all these factors in to account.

How to!
Here is one way to start a good ad campaign and paying as little as possible:

1. Define keywords that are relevant to your website or create relevant webcontent first (to lower minimum bid prices). Group them so one ad text and landing page can cover the entire group. This probably means small groups.
2. Search your own website for each keyword and optimize a high scoring page as the landing page for that keyword or small group of keywords. Do extra linkbuilding and wait for organic ranking first if the minimum bid price is too high.
3. Enter the keywords as exact match and broad/phrase match and supply as much negative keywords as you can think of. Make sure you’re able to measure the exact typed-in search phrases in your web analytics.
4. Create an ad text that is different from all the other ads ranking for those keywords. Click-through-rate is mainly based on: Offering what one searched (ad relevance), offering something seemingly better then the other ads (ad specs) and sticking out from the crowd (ad call-to-action and ad position).
5. Enter the landing page URL exactly as the page scoring in organic search. Without any URL differences you could use to recognise adwords visits.
6. Find out/guess what position other than the first one is highly visible with your ad text sticking out from the surrounding ads. Aim for this position when setting your maximum cost per click. Start bidding high, but when your historic click through rates increase you can lower your bid to keep the same ad position. Adwords also has a prefered position setting, so use this to automatically lower your CPC in time.
7. Set the daily budget high at first so you can do keyword research. If you don’t add all your ads at the same time and finetune before you add another, you can see what keyword potential is available and you get a representative percentage of searches to finetune the different matches.
8. Add the ad group and start by finetuning your search terms. Look at your web analytics and see what exact phrases are used within the broad/phrase match. Enter all of them as exact match, but use negative matches for the ones that are irrelevant for your website. If they are just irrelevant for that particular landing page, create a different ad group with a better landing page.
9. Then finetune your ad text. Enter two or three slightly different ad versions and see what ad text converts best measured in click-through-rate.
10. If one ad is chosen as the winner, reflect that ad text in your landing page. Preferably you should use the same title and description.
11. Once you have almost no clicks on your broad/phrase matches – because you use exact and negative matches that include everything contained in them – then you can activate a second ad group and repeat the whole process.

SEA requires SEO
By doing all the above you get high click-through-rates, high relevance to keep your visitors, low minimum bid prices and lower average cost per clicks. Most SEA agencies do only CPC bid management to increase the effect of your ad and lower the cost. It is about time they include ad and landing page relevance in their routine too.

Indicators of ad and landing page relevance are the same as indicators for organic relevance. Most search engines just don’t use all their organic spam filtering on quality scores because there isn’t that much landing page spam yet. As more sites start influencing relevance in the quality score artificially search engines are bound to add more spam filters and SEA pricing will become SEO work.

What is keyword density?
Keyword density is the percentage a keyword is of the total amount of words, a specific piece of text or code, or a percentage of all the characters. Besides density, the keyword prominence is also an important factor in keyword placement. Prominence is mainly how far the keyword is from the beginning of the text or other important spots within the code.

Is there an optimum keyword density?
Yes, there probably is! But it is a different density for each search engine, each HTML tag, each keyword and the amount of words. The exact percentages change all te time. An optimum density is more a minimum and maximum density to work between, then an exact number that is perfect all the time. What the minimum and maximum percentages are depends on what numbers you use to calculate them.

Calculating keyword density
There are several methods and tools to calculate keyword densities. In my opinion the best way is to take the total amount of words between a specific region or HTML tag. The most important densities are: Percentage of total words within any tag; And percentage of total words within the body tag.

A good tool to calculate keyword density is Ranks.nl (http://www.ranks.nl/tools/spider.html). This tool gives you many options to measure and compare the numbers. The optimum range I will give later in this article are ones you measure with ranks.

What should the numbers look like?
Exact numbers aren’t very important for your ranking. Even the numbers I give here are just indications and a page that has entirely different numbers could rank just as well. The closer the numbers are the better, but in the entire algorithm it is just a small factor.

• The total number of words shouldn’t be too high or too low. Optimum amount for the percentages I give below is between 300 and 600 words. Use as little code besides text on your page because it could all dilute the importance of the main keyword.
• The main keyword should always be represented in the page title (title tag). Other important places are: URL, headers (h1), linktext, bold text, italic text and alt tags. Within these tags the main keyword can and should have a very high percentage.
• The main keyword should have a higher percentage than other words. It is not the percentage that counts, but the focus on the keyword compared to other words. In my opinion this is the main ingredient of good page optimization.
• A percentage higher than 3% and below 10% of the total wordcount is best, but I often keep it below 5%. All unimportant words should be kept under 3%. Avoiding unimportant repeats is more important than repeating the main keyword.
• Stems are the basic words that keywords are made of. A keyword can contain multiple stems. All the keywords posting, poster, posted, postmodern all contain the stem “post”. Keywords or stems that are included in other words are counted in the algorithms of some search engines. Keep keyword repeats including stem repeats below a certain spam threshold.
• If you repeat keywords too often, use synonyms so the theme of the page is still strong. The theme of the page is almost as important as the keyword itself. Some search engines are good at theme destillation (guessing what a text is about) and some don’t do it at all. In the future it will become more and more important.
• The “one page, one keyword” strategy works best, but for words with little competition you can always combine multiple keywords in one page.

My opinion:
Focus is everything! Focus on keyword related text and prevent unintentional focus on another topic. Focussing is done in repeats and accentuated code parts (title, bold, etc.).
The theme of pages (internal and external) linking to a page is more important than having perfect densities. So focus on theme on a wide scale and focus on keyword on a page scale and in linktext.

Please comment because this topic is something opinions vary widely on. How important do you think keyword density is?