Word Order in phrase important?

[dylanthered]

I'm trying to pick the right keywords for my various pages. I've used Overture, and WordTracker. What I'm confused about is that Google gives different results depending on the order in which you place words in a phrase or whether they are singular or plural. Do any of these tools distinguish between word order? For example, if Overture says I get 591 results for the word "dallas charter bus" does it mean just that or does it combine all searches including "dallas bus charter", "charter bus dallas", etc.

[orion]

Not that I aware of.

This is where most keyword discovery toys and keyword research services out there fail to provide to customers.

Back in SES, San Jose, we have a private party at the Marriot vicepresidential suite. I meet there Andy Mindel (Wordtracker) and one of his programmers. I talk to the programmer and presented him a problem need to be addressed. I also discussed this with Mike Grehan. They were taken by surprise with the following I presented to them...

When you discover terms from these toys, you get search volume data. Fine.

But let say you get 10,000 search volume counts for k1 k2. Here is the question: Which fraction of those were entered by the users for the EXACT sequence "k1 + k2" and which fraction were queried for the terms in no particular order?

Which fraction is for k2 + k1? Which is for k1 + k2? And for the exact sequences: which is for "k2 + k1"? Which is for "k1 + k2"? Which are for phrases (k1 and k2 delimited by a space)? Which are the results of ignoring delimiters placed within the k's? (e.g. k1-k2, k1_k2, k1 & k2, k1 "and" k2, k1 "of" k2, k1 "in" k2, etc...?)

To top off, which of a reported search volume sequence is the sequence more prevalent in a search engine database? Wordtracker and Overture do not tell you this information. These and similar services simply present to users a global search volume number collected from specific databases. Unless you are a savvy keyword researchers, blindfold relying in these toy counts can be reduced to a case of garbage in, garbage out.

For a three-term sequence, above, the scene is much more complex (4 c-indices, each with their corresponding permutations and EXACT sequences). For 4-terms and above, forget about it! Trying to combine metrics from disimilar keyword research services as some in the industry have been proposed (SES, NY)? Forget about it as well.

The above problems are extremely important when we do ontology studies, phrase matching and on-topic analysis from search engine metrics or keyword research. A mere search volume count or looking at dictionaries/thesaurus -as some proposed at the SES, San Jose, keyword research track- is in my view a dumb strategy since this does not answer the questions of word relationships in a lexicographical tree, nor address the issues of which on-topic terms and sequences your visitors target.

Orion

[dylanthered]

Thank you for that thorough reply. So I was right? This is a big issue for everybody, not just me. In my case I have to optimize for local search terms. An example of my problem is choosing three keyword phrases if I don't know the importance of word order and each phrase has three words.


Example:

If my phrase were "Dallas limo rental" then I would need one page for:

Dallas limo rental
Dallas limo rentals
limo rental Dallas

Then another page for other phrases including "limo rentals Dallas". The permutations get crazy. I can only guesstimate using the tools of what to try. I guess the more pages I make the better.

- Dylan

[seobook]

Originally Posted by dylanthered

If my phrase were "Dallas limo rental" then I would need one page for:

Dallas limo rental
Dallas limo rentals
limo rental Dallas

Then another page for other phrases including "limo rentals Dallas". The permutations get crazy. I can only guesstimate using the tools of what to try. I guess the more pages I make the better.

not needed. you can mix anchor text to hit different variations and permutations. I would not suggest making a half dozen near similar pages for each core term if it is a good legit site you want to keep longterm.

you also can set up a test adwords account to see what versions are the most common...although that might cost a good bit. keep in mind that you have to account for search quality as well, and search volume may not be the only thing you should look at.

using some other related words & phrases on the page like
-for hire
-cost
-luxury
-cheap
-affordable
-with pool and bling

may also help bring in hyper targeted traffic that converts well.

don't forget your modifiers.

[orion]

Originally Posted by dylanthered

Thank you for that thorough reply. So I was right? This is a big issue for everybody, not just me.

Indeed it is. Because of the above reasons, I don't believe in keyword research or web analytics that relie on these services.

In my case I have to optimize for local search terms. An example of my problem is choosing three keyword phrases if I don't know the importance of word order and each phrase has three words.

You mention you need to target local search terms. Then you may want to think about identifying/prioritizing the scope (state? city? zipcode? street? radius?, common usage in the area, etc) and about what local users search for.

To identify the importance of word order I recommend an on-topic analysis for the document. This would extract useful combinations per documents.


Of course, you can always try cutting corners, sloppy approaches and brute force term transposition but I don't recommend those approaches.


Orion

[dylanthered]

Yeah, and someone said something about WordTracker. WordTracker doesn't solve the problem either.

What about using Pay Per Click and configuring the settings so they only came up on very specific search entries. Would that help with word order and plural permutations?

Can you explain a little what you mean by an on-topic analysis?

Thanks,

Dylan

[orion]

Originally Posted by dylanthered

What about using Pay Per Click and configuring the settings so they only came up on very specific search entries.

You can try that but notice that a recent industry study and the usual PPC vs SEO thing is currently under fire: http://forums.searchenginewatch.com/...2408#post62408

Can you explain a little what you mean by an on-topic analysis?

Sure. Check the SEW thread On-Topic Analysis


Orion

[Marcia]

I would not suggest making a half dozen near similar pages for each core term if it is a good legit site you want to keep longterm.

Good advice back then, and from what I've been noticing, even better now. I've seen sites that did that dropping down in Yahoo rankings and too_similar pages ending up in Google's Supplemental index.

Anyone agree that it's better to diversify using synonyms on pages/sites to pick up added traffic?

[DanThies]

While word order does matter to an extent, it's not always that important. I'm sure I've posted before about E/F ratios and "implied quotes," where we see strong indications that search terms commonly used as phrases are more likely to see word order as an important factor in search results.

In any case, a few things should be clarified:

• Wordtracker does not re-order search terms, as Yahoo/Overture does. You can use the multisearch tools on Wordtracker to find out how many searches are done with different word orders, and do the math.
• You can use an E/F ratio (read this forum) to get an idea of how important word order will be in search results / rankings. You don't have to believe this, and I won't debate it with you.
• Keyword Discovery has an issue with this because so much of their data comes from PPC engines, which do reorder the words in a search term. Trellian has addressed this at the HR forums, and may offer an organic-only database some day.
• Using the most common word order on a page, combined with other word orders in internal links, has been shown to yield good results. Link text is also handy for targeting misspellings, local terms, etc. and generally any time you need to "overload" a page with search term variations.

[earlpearl]

From a competitive perspective it is critical and vital.

I run a local business/site and compete for industry keywords within a two state one city area.

One competitor in the other state has a url that is www. statenameservice .com

Try as I might...with lots more bls, including backlinks with specific anchor text including significant relevant local backlinks and content that places me within that other state, for the phrase...state service....the other site ranks first in google and my site ranks 2nd. Reverse the prhase with service first and state name second my site ranks first.

It means a difference in traffic and business. Its important. Its not the be all and end all but it is important. As a long term business, I can't tell you the number of customers that called us over the years with the first listing in the Yellow pages, versus the second listing.

Some potential customers will only go to the first listing they see. Others will call a number of vendors. It does make a difference.

If you are going to chase a variety of long tail and compelling keyword phrases (as described above) give the interior page names the keyword phrase. Get those titles in the interior pages.

On the geo side, various vendors suggest that in some industries people use the location name first and the service second or vice versa. Based on about 2 years of kyword data and comparing traffic for the jurisdictions where we are ranked first for either case...I see a slight orientation toward service first, state name second...but not dramatic. When I take a look at the data for the state where we are second...we just don't get as much traffic for the phrase with the state name first. (Its better to be ranked first.)

Regardless of how the various keyword tools work....you need to keep tracking this over time and make the effort toward higher rankings.

Also experiment with PPC and PPC rankings.

Lots of luck.

Dave

[orion]

Originally Posted by DanThies

Using the most common word order on a page, combined with other word orders in internal links, has been shown to yield good results. Link text is also handy for targeting misspellings, local terms, etc. and generally any time you need to "overload" a page with search term variations.

This Dan's advice is correct and on target.


However, two problems with most of these commercial keyword research services are

1. database fusion = combining counts from dissimilar databases is equivalent to adding unnecessary noise. This noise is non trivial and can mask keyword usage patterns

2. ordering = here we need to think in terms of the effective ordering of terms, or technically speaking, know the difference between literal word sequences and relative word sequences.


My experience with SEOs is that they are notorious for thinking in terms of literal word sequences, overlooking the later (relative word ordering), without no mention about relative word ordering within shifting windows (shingles) of a given length, L, -important with Google's query-sensitive duplicated content algorithms. Evidently, if one study these algorithms, it becomes crystal clear the importance of relative word ordering.


In 1, is like trying to get users analytics from K-Mart to make predictions for Wal-Mart. Pure non sense.

In 2, relative ordering is implicit in EF-Ratios, and goes back to the notion of what is an EXACT search, which with Google not necessarily is a search for phrases.


To review this, an EXACT search of the form "k1 X k2" will retrieve documents in which


X == space (as in phrases)
X == a dash (as in hyphenated terms)
X == a period, question mark, exclamation mark (as ending one sentence with k1 and starting the next one with k2.
X == delimiters surrounded by spaces (as a pipe, a dash, a bracket, etc)
X == punctuation specific delimiters (as a comma, colon, semicolon, apostrophe, etc)
X == stopwords (as the, of, a, for,..., etc).
X == valid words to be ignored, but not necessarily being filling words.



Evidently an EXACT search is not a search for literal sequences (phrases). This has important complications when we calculate EF-Ratios or when we optimize since the metric is not a measure for phrases but for the fraction of relative ordering within the FINDALL answer set under consideration.


When keyword research must be correlated with seasonal trends or in-progress events, the best way to datamine keyword usage is by monitoring EF-Ratios over time (Temporal Co-occurrence) A study published a while back that discusses Temporal EF-Ratios and that put into question the validity of keyword research services out there is available here. See Temporal Co-Occurrence: How does a Developing Event Affects Search Results?


Orion

[earlpearl]

For a local business its a very important issue. It is worthwhile to strive for top rankings for phrases that combine a geo location and business service/products. Its often achievable, contingent on the level of local competition. At times the competition is significant, at times it isn't. That means that first rankings for those phrases in either order with geo name first or service first are achievable.

While I shot off a quick response, your comments and those of others made me look harder at my own web traffic. I looked at my log analyser going back to Feb last year. Last year during the Super Bowl update of 2005 Google started showing local business sites for quieries like limo service Dallas. Previous to that the top ten rankings were dominated by mega sites like Amazon and a great variety of directories, all with tons of backlinks but relatively little local relevance.

My own business site is well optimized and shows welll for both the relevant local terms (business service/geo descriptio) and for the industry terms (no geo description). After the Super Bowl update we started seeing competitors show in the relevant rankings.

For my site I dug through the top 200 keyword phrases (so far) looking at the #1 industry term and relevant geo phrases. There were 50 such keyword phrases with combinations of geo term/industry phrase. (never looked at this before).

Somewhat over 2,000 phrases had the service first and the geo description second and about 500 were reversed with geo description first and business term 2nd.

The site/business primarily competes within 2 states and 1 city. So there are a tremendous number of variations on the combination of keywords and geo descriptions with full state names, initials, combinations of the city and either state, regional descriptions, town and county descriptions. They are all worthwhile to add into your content to pick up traffic and conversions.

BUT, that is not universal. Web Real estate people have told me that more searches are conducted with regional/geo description first and real estate second (i.e. Dallas Real estate, Dallas homes).

It is different for every industry. After reading the above I'm not sure that I understand which tools do a better job of differentiating between word/phrase order.

I do think all comments above by Orion, Dan and SEObook are relevant and helpful.

I've also found that getting the word order in the url and/or interior page makes a big difference for rankings.

Incidentally I found a derivative tool off of overture that does give some additional insight into phrases and word ordering for local services;

http://www.rightscripts.com/keywordresearch/

Try it with an industry phrase that would be highly relevant for local searches. It gives further breakdowns per city/region that might add some insight into your efforts.

As seobook referenced it is helpful to have compelling additional descriptive phrases connected with your service and also important to grab at the long tail secondary terms. Then these need to be caught by the SE's with the relevant geo descriptions.

Many sites combine a bunch of affiliated keyword phrases on one page. If your service can be described 2 or more ways it can work to have the different descriptions on one page with the relevant geo descriptions. The site can be picked up by the engines.

On the other hand you can and should add pages for variations on the service and/or products. Limo service could have a page for proms, weddings, business, etc. All of it helps capture the long tail. Capturing the long tail and variations can be every bit as important or more than word order contingent on the business/service.

Plural and other types of stemming are treated differently in my experience. Google does an effective job for stemming for singular and plural. Y and MSN don't do it as effectively suggesting that you need different pages for plural and singular. (but that could vary service to service and industry to industry).

Actually, I love questions like this as it forces me to do some deep research to see if I can better understand patterns as to what works.

Good luck.

Dave

[orion]

Here are several conclusions from the quoted article (emphasis added):


"While our experiment was quite preliminary and limited to monitoring document volume over time, search volume can also be monitored in a similar way. This might interest companies that provide search volume services.


Currently some search engines and marketing firms provide these type of services. For instance, Google publishes Google Zeitgeist (12). Overture (acquired by Yahoo!), Wordtracker, Trellian and others provide a variety of search volume services. Monitoring results from these services over time should allow one to extract important word patterns and seasonal trends.


However, when using such services one should be aware of what the reports actually account for. One may want to consider the following:


Are these the result of fusion; i.e., of combining results from dissimilar databases?

Are these the result of mixing; i.e., of counting hits for a term regardless of whether the term was queried alone or as part of a phrase?

How do the reports discriminate between query modes? If so, which fraction accounts for users searching in EXACT or FINDALL?

How do the several containment and co-occurrence responses evolve over time in the websphere or blogosphere?

When, why or how do changes in overall trends occur?

During which time interval the measured response ended contaminated with a significant amount of noise?


If the reporting service combines search volumes from dissimilar databases or mixes search counts more likely that will be detrimental to the research. Unless stated in the report, a keyword researcher will not know which fraction of results comes from which database or how partial counts are affecting total counts.


On the other hand, if a keyword reporting service cannot discriminate between query modes, then the reports might account for what users searched, not for how users actually searched. Under such contaminated scenarios, trying to extract word patterns or to correlate searches with user's interests or search behaviors might as well compound the noise.


Conclusion


We have presented a preliminary experiment aimed at addressing how an in-progress event might affect search results from commercial search engines and when these deteriorate significantly. The experiment was conducted in the websphere and blogosphere using both FINDALL and EXACT query modes. Several containment and co-occurrence responses were tested, with the EF-Ratio outperforming all of them.


Our results suggest that FINDALL -default search mode of most search engines- do not provide a reliable analytical response for conducting temporal studies shaped by developing events, although this may be possible in the blogosphere to some extent. The experiment questions the validity of keyword research reports that do not discriminate between query modes or database results."


It is clear that the noise I describe is compounded by many of these keyword research services and by those building tools/services based on these.


Orion

[orion]

The quoted article was first published last November of 2005. Back then there was no way of comparing document volumes with search volumes or any two search volumes in Google.


All that is possible now (up to some level) with the release few days ago of Google Trends. Despite its many limitations, the service provides at least a way to look back in time and compare search volumes in the websphere and newsphere.


I've added an appendix (Appendix A) to show how one could do this type of analysis with Google Trends.


This is a plus, Google. Congrats. The next thing is convince you to add a signal for direct measurements of EF-Ratios signals in time (temporal co-occurrence).

Orion

[identity]

I think you have to approach geo positioning based on either your topic or your user target.

Real estate... location, location, location... literally, so it makes sense that users are thinking of location almost first and foremost and would be more inclined to lead with their search.

Topic and location... outside of real estate, again, is it a highly location sensitive topic... obviously there is a lot of gray here.

So moving on,

User target... in this case, I think it may depend on how web savvy they are in relation to the above. I suspect that a less web savvy person starts with their topic, then when realizing that it isn't delivering what they want, perhaps limo service but for LA, New York, and maybe even Podunk, rather than starting over, they append a location... state, city, whatever they feel is appropriate.

On the flipside, a very web savvy person, from previous experience, may know that "limo service" might be a fine search if they were seeking general information, but if they were seeking "limo service" for the Dallas area, they will initially start there. It may still be a question as to whether they prepend or append "dallas", but I'm guessing that "dallas limo service" represents more of the experienced users and "limo service dallas" represents more of the less experienced users.

all opinion of course.