Forum Moderators: open
For all the luddites and technophobes out there who (inexplicably) still want to do keyword research, don't despair! You don't even need a spreadsheet. Two books will suffice ;)
I'm not sure that 'widgetising' is going to help me with this example...
Let's look at the example of Mr Widget and his Widget Workshop. My first stop is at a dictionary. You can cheat and use an online one if you must.
Here's what I got for 'widget':
noun (informal): a small gadget or mechanical device.ORIGIN - perhaps an alteration of GADGET.
I now know that people might also refer to widgets as 'gadgets' or 'mechanical devices', and widgets are usually small. There's a suggestion that if my audience is likely to be more formal, gadget might be appropriate. Time to check a thesaurus (which sends me over to the entry for gadget):
concern, contraption, contrivance, gimmick, jigger, thing, widget
The presence of 'thing' is a concern, since it implies my potential customers might not even be sure how to describe their widgets! I think I'm stuck with 'contraption' as the only keeper from that list.
This might be the right time to check some of those new words in our handy reference books, however I'll spare you the gory details.
Just as important, of course, is the specific offering of Mr Widget, who repairs widgets. You know where I'm going with this ;)
The dictionary and thesaurus give me a whole host of other words, including fix, maintain, mend, overhaul, patch, revamp and restore. A pen and paper is all that's needed to make some notes and put the words in the order Mr Widget likes best (he knows his widgets!):
Nouns
Verbs
Adjectives
By my reckoning, there's 20 simple combinations of noun and verb alone. Once we account for adjectives, plurals, parts of verbs, word order and so on, we've got everything from 'widget repair' to 'restoring small contraptions': many thousands of potential keywords. Mr Widget and his customers can give us some insight on which phrases are the best to use to describe his services. Not a bad result from two dictionary entries and two in a thesaurus!
So now you want to know which combinations are searched for most often? OK, OK, I give in, although there are plenty of other places to look for useful information. Next time I'll go back to a PC/spreadsheets and post about statistics ;)
But it was a useful exercise, wasn't it? :)
The presence of 'thing' is a concern, since it implies my potential customers might not even be sure how to describe their widgets!
noun (pl. thingies) another term for THINGUMMY.
This is starting to get complicated ;)
Of course, aside from specific words and phrases, you don't need a computer or a book to know the general words that will apply to most keywords: action words (like buy, order, purchase) and so on, the obvious qualifiers like regions (e.g. UK, Midlands, London) and other words like company, firm and supplier.
I think my 'no-computer' list may have become unmanageable ;)
oh the "sniglets" we can come up with :)
Word-stemming has become severely important to ranking well for selected keywords/phrases.
Some time back, Marcia (former moderator of Google search News) posted a few links to the Google patents.
[webmasterworld.com...]
They are well worth reading.
At first blush, one would think that a good start to discovering what Google might be looking for re: related words/phrases would be to examine No's 1,2,3,4, & 5 in the desired SERP's, click on
"Similar Pages" below each, and see what's there.
Alas - not very reliable. Why, I can't say; But after months of trial & error...
I made some pretty useful discoveries.
Why? Because the few pages in the 95%/5% website had good internal links to adjacent pages
full of related words/phrases.
E.G. A website about "telephones" is well served to include words/phrases such as 'DTMF' /
'dial tone' / 'handset' / 'cords' and other such words/phases that aren't in the typical layman's
"telephone vocabulary".
To supplement this, may I also mention:
Moreover, Google would often list this pop-up page, (provided it was a full-flavored HTML page with good copy but with it's size trimmed by the pop-up command's dimensions) as the 2nd most important page in the site "about" the key phrase -
However, the pop-up page needed different <title> and <description> text;
Again, using word-stemming.
A good question is: What's all this worth?
After many months, and following the above somewhat didactical policies, I drove three different websites from <50 to Top 5 spots in the desired SERP's.. and some of these SERP's had dozens of relevant competitors, with better PR!
.
Just my two shakes of flavor..
discover what people are actually searching for
Amen! and Amen! again..
I can't count the number of site owners who've told me..
"We're not seeing enough traffic from Google, what's wrong?"
I look at the SERP's they are in, and 80% of the time, they are
'selling' terms that
a) Nobody is even aware of.
b) Are way too industry-specific.
c) Aren't covered for mis-spelling or truncated versions
d) Aren't backed-up with terms Mr. Consumer would be using
when "conceiving" what they do/make/sell -
And therefore, would in fact be searching for.
:)
.
The dictionary and thesaurus, a book on the subject matter, coupled with sample searches, pulls up the majority of short and long tail terms, for me.
Put your mind in the place of a prospective searcher.
Go for phrases and terms that people would naturally type.
The point of keyword research - for me anyway - is to discover what people are actually searching for.
Go for phrases and terms that people would naturally type
While these two goals should be mutually inclusive, there can be an element of artificiality (as engine suggests) to many keyword databases that can be unhelpful.
The other problem with the 'tool-based' approach is that it can dilute focus from keywords that are likely to drive the right sort of visitor - i.e. those that result in sales or other desirable actions.
A useful approach for lovers of statistics (i'm one!) is to start with 'natural' keywords, and then collect usage data on those - cutting out much of the unreliability of external databases. Perhaps that's worth a post in itself.
The biggest problem with bought in databases is that the information is skewed by people checking on their own sites.
>How about reading niche forums and sites like Yahoo and Google Groups?
>How about going to the library (remember those?) and skimming through periodicals that have spent a ton of money on how to use language in their publications that will appeal to their target audience?
>How about going to a "boring" textbook on grammar to refresh memory on what adverbs and adjectives are?
Anyone think users don't use adverbs and adjectives when they think and when they search? Think again!
How about going to the people themselves who we would actually want to reach?
periodicals
[edited by: TonyMc at 2:03 am (utc) on May 9, 2008]
If you fine-grain it, it becomes difficult because you are assuming the SE's are using a dictionary-thesaurus platform to fuel the algos.
Language is fluid, however, and if the workings of Latent Semantic Indexing (LSI) is anything to go by, there is no need for a traditional dictionary/thesaurus... nor is it necessary to "know" what words mean.
So basically, you can use a thesaurus, by all means (it will help), but to use it for pure SE economics, don't waste your time: the tilde operator (in Google) shows the difference between academic word-relations, and real-world, "actively monitored", relations, computationally derived.
you are assuming the SE's are using a dictionary-thesaurus platform to fuel the algos
Perhaps it came across that way, but I meant something quite different. I'm assuming that a writer can interpret dictionary/thesaurus entries much more effectively than any current search engine. A search engine relies on a mass of user-supplied data (much of it inaccurate) to approach the same task. With rather mixed results.
the tilde operator
It tells us a lot about Google's current 'understanding' of words. But it also highlights a lot of misunderstanding. I reckon you could get a better list of related words by asking the nearest person to you at the time ;)
We're coming from very different sides of the debate. I think search engines are playing catch-up when it comes to keywords - that's why a concept like LSI even exists, and to some extent is a major reason something like SEO is even necessary.
It'd take a week to write an article employing the OP's theory
I remain a single-finger typist, but a week would be stretching it a bit ;)
It tells us a lot about Google's current 'understanding' of words. But it also highlights a lot of misunderstanding.
That's my point, yeah: real-world.
It's a douible-edged sword - I've seen quite a few synonyms in Google that my thesaurus missed. I've seen the names of manufacturers "linked" (correctly) to abstract brand names... by abstract, I mean word(s) that don't "officially" exist.
The power of computers is often their ability to make connections (and groupings) we don't immediately see.
I think search engines are playing catch-up when it comes to keywords
Catch-up to what, though?
Catch-up to what, though?
To understanding words; translating a user search to a relevant result.
Every internet search involves a person who usually has a very clear idea of what they want to see when they undertake a search. They'll probably type two or three keywords to try to describe what they're looking for. But if you ask them, they can tell you all kinds of other information about what page they are looking for, including massively complex associations between different words and topics.
Search engines try to close the gap by analysing patterns of data - a very reasonable approach because they need massive scalability - but certainly isn't perfect ;)
I think this is a key advantage that you as (for example) an SEO have over a computer program. Of course, you can also look at how the search engines approach things and combine the two: something that a search engine will never be able do.
Every internet search involves a person who usually has a very clear idea of what they want to see when they undertake a search. They'll probably type two or three keywords to try to describe what they're looking for.
So given the limitations imposed on the SEs, by us mortals, it could be argued that they - more specifically Google - perform an amazing job... maybe it is we, that needs to catch up; could you do a better job, manually? ;)
But if you ask them, they can tell you all kinds of other information about what page they are looking for, including massively complex associations between different words and topics.
And this is where the "advanced" operators come into play: daterange, subtractors and the like, for instance. I think we're now talking about the longtail and corpuses, and stuff.
Again, though, human language itself is limiting - some languages more so than others.
Ask five people to describe the same room in three words - You get the picture, I'm sure.
