Keyword density experiment

I was thinking of a way to prove/dissprove keyword density does not effect SERPs. The experiment though will need the approval of Seochat moderators though.

If we were to start five identical threads in this forum with the only difference being the first (and only) post is about 100 or so words with almost exactly the same content except for different keyword densities (useing a unique keyword) and than see which thread ranks best.

Than repeat with new text and see if the same is repeated. The speed with wich google indexes this forum we should be able to see the results almost immeadatley.

(Or maybee sligthly different words in the text (other than the unique word) but exact same number as the other txt shoudn't matter to KWD test)

The threads could be delated within 1 day.

Comments, suggestions, improvements..

It sounds good to me. Hopefully we can then prove that keyword density is a dead duck and put the issue to bed once and for all.

My only query though is with the methodology. To my mind it wouldn't necessarily disprove that keyword density is an issue: the results could be interpreted as being that the top ranking thread shows that its keyword density is optimal, 2nd ranking thread that its keyword density is the next optimal, etc. I'm not sure that it would convince advocates of keyword density to change their mind.

Having said that I'm struggling to come up with a better method.

Maybe an alternative approach would be create a test thread based on normal writing patterns (ie low density) and without unique keywords, see where it appears in the results for the test keyphrase, then gradually increase keyword density around that phrase. If SERPs don't increase as a result, that IMHO would be better to prove that keyword density is not a factor.

My idea is if this experiment is run say three times and KWD of the first page in serps changes each time it shows it does not matter. If the same KWD of each test always gives the same first result than it would support that KWD is a factor in rankings. The beuaty of using this forum is the speed of indexing we would see th results with hours.

IMHO, this would be a better option.

Ah right, with you (sorry, it's a Tuesday - takes me until at least Thursday to fully engage the brain ). That would certainly appear to make for a sound experiment. I'm for it!

The first thing is "dup content"... SEOChat's margins and menus are a major portion of repetitive realestate - to add identical thread (with only a marginal change of wording will get all thread penalize quickly but 1... obviously a false KWD positive is the outcome.

IMHO the experiment doesn't warrant any serious consideration.

KWD implies "what level I can keyword stuff without losing ranking ground".

You can color that different shades of blue or pink but really - if any experiment suggest that 50% KWD was the PROVEN best... what would happen...

The one thing people who push KWD never consider is percentages above your suggested density... what occurs?

For argument sake let's say using 10 phrases for a 500 word page is suggested as the best keyword density... search engine don't PENALIZE above that - you have just reach the cutoff (like a 64 character title element cutoff in Google) and no more benefit can be had... (NOTE: how many people actually believe if you go to 70 characters for a title element Google penalizes you? NONE I HOPE!)

So using 11, or 12, of 20 phrases on the page produces no more or no less than the impact at 10.

So what other variables do we use in best SEO practices that impact on the pages ability to rank - that also impact on keyword density?

Title Element - if your phrase isn't here no amount of potential rank KWD elsewhere will have "the best ranks you can achieve".

Page Title in H Tags - if your phrase isn't here no amount of potential rank KWD elsewhere will have "the best ranks you can achieve".

Links from subordinate pages - assuming you link to subordinate pages of the same or similiar topic suggests you are using at minimum partial matches similiar to the effect of stemming... to illustrate better if you use the words search engine in conjunction with Marketing but your seeking a KWD for search engine optimization the short stemmed search engine still impacts even if Optimization isn't used - so we can be reasonable certain that 1 or more links on a page with use the keyword or partial keyword.

That's 3 uses that MUST OCCUR WITHOUT CONSIDERING KWD.

Other Possibilities

Alt="" an image somewhere on the page uses the keyword.

Copyright Statement - domain name or business name

What else?

OK so we have 3-5 uses BEFORE WE START WRITING KEYWORD DENSITY BASED COPY... so what do we have here without a single word written to maximize KWD...

BASED ON 500 word-page copy:

If the phrase is a single word - KWD - 0.6% - 1.0%

If the phrase is a two-word phrase - KWD 1.2% - 2.0%

If the phrase is a three-word phrase - KWD 1.8% - 3.0%

This is before we start even writing... so for the sake of argument let's advise people it is a damn good practice to:

Use your phrase in the Title Element

Use your phrase in the Page Title

Use your phrase in internal links to the page

...AND WRITE FOR THE READER (BUT ADD THE PHRASE AT LEAST ONCE IN THE PARAGRAPH TEXT)

and their KWD is:

If the phrase is a single word - KWD - 0.8% - 1.2%

If the phrase is a two-word phrase - KWD 1.6% - 2.4%

If the phrase is a three-word phrase - KWD 2.4% - 3.6%

... what's the upper cutoff?

Does it really matter?

The best KWD advice is IGNORE IT! ...without experimentation it takes care of itself with proper crediting, copywriting and wordsmithing.

Absolutely, but isn't that the point of the experiment: to prove that KWD within the main body text is not a factor in SERPs.

At least then we can point to it next time someone asks about KWD and back up the statement that it has no bearing on rankings.

Yep, there is a possibility that several threads, each very similar, could cause some to be filtered for duplicate content which would be an issue and skew the results. So the method needs to be refined, but I'd still like to see a quantifiable experiment done on this, as no amount of logical discussion and anecdotal evidence will be enough for some to abandon the notion that increasing the number of instances of the keyphrase on a page is going to improve their results.

Yep.. Thats it. As for the dup content issue its the amount of words that matter not what they are. So if say there were 100 words in each thread and many were changed with same amount of letters this should avoid duplicate content issue.

Fathom I agree with you about KWD its just that this is an experiment forum and I thought this would be an easy simple experiemnt.

I am not 'pushing' this though if others do not think its worthwile fine. It would need the support of moderators to run. The reason this is easy to test is the speed of ranking and indexing this forum gets. This removes the other variables. We can also run the experiment over a number of times. This is the perfect place for a controlled test to prove that KWD is not a factor in rankings.

Fair enough...

don't use any phrase in titles or links 'to or on' the page

... you must repeat this multiple times (not just once - not just a single set of densities) since "just random results" doesn't show "any randomness" if you only have a single occurance.

I conclude now - your total observations will produce random results - e.g. no KWD is better than any other.

If it doesn't then I vow to eat my own keyboard

One other thing...

I assume that the bases for this experimentation is to measure "tangible results" that KWD improves ranks... not just to see if there is any movement what-so-ever. (Analogy an ant WILL exert a force on your car making it easier for you to move it... but the ant's force isn't really a tangible gain).

Ranking difference between #999 to #900 is a ranking improvement but that improvement "isn't tangible SEO results"... such an improvement in SEO tangible ranks position #30 and above will never be seen.

Based on the competitiveness of "nonsense-words" so you can actually see and make observations are not meaningful observations in themselves... they don't help to support a conclusion that KWD aided enough to warrant it consideration in SEO practices.

I would think then to guage this appropriately you need to use 100 other pages to produce a reasonable measurement of KWD ONLY in comparison to:

the phrase only in a Title Element
the phrase only in a Page Title (H1, H2, H3, H4, H5, H6)
the phrase only in a Meta Description
the phrase only in a Alt Tag
the phrase only in a link to the page

ETC...

If all of these single use non-KWD-baseline variables all appear ahead in results - it will provide a rational to use the phrase Title Element, Page Title with any of H1, H2, H3, H4, H5, H6, in a Meta Description, in a Alt Tag, in links to the page but the value of any KWD itself - is a non-issue... I would think.

You could mix and match of the above "other variables" to come out with 100 variations - and I would guess that a single visible phrase on a page will beat hidden stuff like alt and Meta - but when compared to the other variables - KWD only won't be in the top 10, top 20 or even top 30 positions... and these are the only positions that matter in SEO... so in comparison KWD would be a SEO consideration...

I would think.

This is a good experiment. I suggest we need to decide on the final flow of the experiment and the data table.I suggest below:

Experiment conditions:
1. Three experiments (one for each: one word,two words, three words phrase)
2. Each experiment has five threads.
3. Each experiment is replicated twice.
4. Keyword occurrence is only on the body content/text and not on other areas.
5. Readability statistics score for each content should be similar,to prevent biasing of keyword density experiment to a high quality written content. In short, quality of the each content should be comparable or similar but using varying occurrences of the targeted/optimized word.
6. No duplicated content as it might bias the results by google filtering.
7. Content in each thread should focus around the keyword optimized (same theme), to eliminate google ranking bias in terms of keyword relevance (not considering density).

Experimental flow for each experiment
Step 1. Decide on a word/phrase to optimize.
Step 2. Make five threads with different content but the same total number of words,but varying different keyword densities.I suggest the following occurrences distribution: Each thread have 100 words each, 1st thread will contain 0 occurrence of the optimized word(0%), 2nd thread will contain 1 occurrence(1%), 3rd thread will contain 5 words (5%), 4th thread will contain 10 (10%), 5th thread will contain 20 (20% max)

Step 5. Gather ranking data after indexation for 1st trial of experiment.Monitor the ranking for the every 8 hours at total of 3 ranking data in a day. (fluctuations of ranking may occur such as increasing or decreasing rankings)


Step 6. Delete the thread.

Step 7. Repeat step 2 with another content but using the same occurrences distribution to get same density.

Step 8. Gather ranking data after indexation for 2nd replication. Monitor the ranking for the every 8 hours at total of 3 ranking data in a day. (fluctuations of ranking may occur such as increasing or decreasing rankings)

Step 9. Put all results in spreadsheet format. (1st and 2nd replication)

Step 10. Analyze the data.

Step 11. Conclusions and recommendations.

Step 12. Publish report here in SEO chat.

This is my suggested plan for the experiment. I am very busy in daytime,but I am happy with my involvement with my ideas above. I suggest this needs moderator participation and some copywriter to make a content of same quality.

Any other plans or thoughts? Thanks.

I was thinking that we should keep it very simple.

5 threads

100 words per thread

in one thread say 1) 1kW 2) 3kw 3)6kw 4)12kw 5)24kw

Run for 24 hours than repeat with different kw (and content)

The order of threads can be altered each time. All threads should be posted simutaneously, and locked immeadatley.

Repaeat over say 5 days.

If rankings of these threads are random over the 5 days than the only variable in the amount of text is kwd this should offer very strong evidence that kwd is not a factor. If one of the pages always comes first this would suggest kwd does play a part in SERPs.

I would propose the text could be from a long article scanned in from a newspaper, magazine, book not currently on the net.

Lets keep this a simple experiment the evidence comes in the repatition.

Further suggestions, comments, etc welcome..

Moderators what are your views. Can we run this experiment. We would need a moderator to volanteer to post the threads and lock them immeadatley. The threads could be delated after 24 hours of being up.

Since this is an SEO Test and Experimentation forum and I implicitly trust and believe all the members who have participated in this thread, so I am willing to volunteer this test if you all believe that this test will really prove anything relating to the subject of keyword density.

However, a few points b4 we all start off with the experiment:

* The content used in this experiment should be unique so that we don't suffer from duplicate content issues.

* The title of the threads would also need to be unique and not like "keyword density test thread 1" ; "keyword density test thread 2" and so on since that might also cause duplicate content issues with thread titles.

* The keywords chosen should not be absolutely obscure ones but should be moderately competitive. We may not really gauge improvements if we choose some obscure keywords coz..for such a keyword even a single occurrence could get a thread to # 1...so the test dies there

Let me know what you all feel about these points.

Another question? Would you all be using the same keyword/s for all the threads or different keywords every time we run the test?? I am a bit confused on this one.

Anyways, I am looking forward to this...but I would appreciate if other Mods and above all JCaputo gives a go-ahead for this. So that we can reach a consensus before running this test.

I'll drop a note to the other Mods and J for their approval and let you all know what they says.

Till then...start preparing things for the test!

What I am proposing is a very simple test. Repeated five times.

The test is keyword density so along as there is the same amount of words in the experiment pages the test is valid. Because if the content makes a difference than KWD is not important.

So 5 threads for 24 hours repeated over 5 days. Each time the previous 5 threads are deleated.

Each day a new keyword that will be in all five threads. The total amount of words in each thread will be 100.

The names of the threads can simply be - one, two, three etc day two - six, seven, eigth etc..

One thread each day has kw 1 time, 3 times, 6 times, 12 times and 24 times. (same keyword different other text in each thread)

The order of the keywords are changed each day in order of threads posted. ie 1,3,6,12,24 day 2 - 24,12,6,3,1 etc

Threads posted at same time and locked.

If KWD is a significant factor in ranking the KWD of the one of these threads should rank above the others every day.

The words can be two real words added together that would not be normally found i.e sport-fairy etc.

The evidence should be in the repatition. As said if KWD is a factor of any significance these kwd that google considers the best should rank number 1 everyday.

Awaiting the yes or no from the powers in charge of the forums.

Note: Remember this is only testing if amount of keywords in page do appear to have an effect on google rankings.