Welcome to WebmasterWorld Guest from 126.96.36.199
Forum Moderators: open
I would like to guess how many times a keyword or keyphrase is searched on internet every month. Let's take 'management consulting' as an example. What would be an acceptable accurate methodology?
I thought I should look first at the estimates of Google's old adwords program: 107.100 per month. Now if I multiply this frequency by the worldwide market share of Google (what is it? 40%?), I would have the Google-based estimation. Let's take 40% market share for Google: 107.100/0.4 = 267.750/month worldwide.
If I take a look at Overture's stats, I get 8009 for june 2002, on Overture US. What is the market share of Overture US? 5%? Let assume it is 5%. Now the Overture-based worldwide estimate is: 8.009/0.05 = 160.180/month worldwide.
Quite different! Of course, I could make an average of both: (267.750 + 160.180) / 2 = 213.965/month.
Suggestions for improvement are welcome. I don't want to make an extremely complicated formula of it. It should stay quite simple and quick to calculate, but still give a good rough estimation.
Of course, monthly stats are subject to variation, but that is another subject.
Certainly I would pay money to have another more reliable source other than Wordtracker, Google Sandbox, or Overture to use for keyword research.
Of course, if anyone else has any nice subjective opinions on this topic I would be interested to hear what they are :-)
For example, for the Market share percentage factor in their formula, I believe WordTracker uses data from WebSnapshot. Using market share from Neilsen, Media Metrix or Alexa might give you a totally different scenario.
Within the word set you are analyzing, there could be a huge skew based on how competitive the keyword set is, and the searches of webmasters checking their sites.
Rather than try to reinvent a square wheel, I'd just pick whichever set of numbers you feel is accurate, and let it go.
You are right: relative frequency is an important information.
But when you talk to a large corporation about SEO, one of the first absolute data they ask is: how many times is 'keyword' searched yearly or monthly. You can't answer 'I can't calculate that for you'. I understand that every marketing boss wants to know what the potential market is he is targetting. That's why a quite accurate estimation looks useful to me.
English is mothertongue to 40-45% of the internet users.
Changes any extrapolations quite a bit, I should think.
Also, if Google has like 40% marketshare in France or Germany, how many searches are done for the french or german equivalent of your keyphrase?
[edited by: heini at 10:15 am (utc) on Aug. 3, 2002]
There have been many attempts, but the problem is that the search engines haven't been willing to cut any deals for the data. WT's biggest flaw is that they pull their data from very low profile sources that primarily push paid listings.
The only way to really improve on what they've built, is to get your data from a better source that more accurately represents the majority of searchers.
Right now, that would be Google, but they don't seem interested in working with a third party.
joined:Nov 11, 2000
wg - I've tended not to trust WT for this reason, and have preferred instead to use Overture with a lot of critical skepticism, and to use the Google tools when search numbers are above their threshold. With WT, I don't like the probable demographics of their sources and the extrapolations that I'm guessing they make. I sometimes wonder, though... am I perhaps ignoring a good source of information?
Given the actual situation:
1. Google has the largest market share but doesn't want to give information about real search frequency.
2. Overture has a smaller marketshare (although all the searches through it's partners do sum up?), and includes 'paid' and 'SEO' automatic searches in their frequency statistics.
3. Wordtracker has based it's frequency stats on small marketshare SE's with paid inclusion.
It looks like a 'weighted average' of the 3 sources should give the 'best' estimation? I'll try that in an Excel calculator...
Thanks to all!