|AdWords' keyword tool versus Wordtracker for Google SEO (Cross Posted)|
Are AdWords keyword counts more accurate than Wordtracker for Google SEO? (
| 4:07 pm on Sep 30, 2002 (gmt 0)|
I have cross posted this question to this section, since I originally posted it to the wrong discussion section (and got no response there). Thanks.
I am currently setting up a Google SEO campaign for one of my web sites.
In the past I have used WordTracker's keyword counts as my primary means for determining which keywords to target in Google. However, for my current Google campaign, I have started using Google's AdWords control panel to extract information as to how many searches are happening per day/week/month on each of the keywords that I believe are significant to my site.
Perplexingly, what I notice is that there are some _very_ large discrepancies between how many searches Wordtracker claims Google is getting on these keywords and the number of searches that Google AdWords' indicates Google is getting for these same ketywords. Often the discrepancies pose big strategic road blocks. For instance, the keyword in my keyword set that Wordtracker indicates is generating the most searches via Google, AdWords indicates is generating nearly the least number of searches.
My question is this: In this kind of situation, should Wordtracker's numbers be trusted over AdWords', or vice versa.
Any help would be greatly appreciated, particularly from someone who has some experience that would suggest which numbers are the more accurate ones, in general.
Thanks a lot,
| 4:18 pm on Sep 30, 2002 (gmt 0)|
There has been some discussion of this in the past... one relevant thread discussing the differences between what's being measured by Wordtracker and what's being measured by the AdWords tool is here [webmasterworld.com].
If you search for "wordtracker" using the site search above you'll find several similar discussions including both AdWords and Overture in comparison with Wordtracker.
| 4:39 pm on Sep 30, 2002 (gmt 0)|
I can only qoute examples.
On one client account we run according to Google's Adwords tools the number of searches that should be found often shows as around 0.1 to 0.6 a day. The reality is that we receive often 5-10 clicks a day for the keywords.
The best thing to do is to use all of these tools as guides, add them to your campaign and within a few days you will know if they are any good. But the beauty is if the words don't get searched for then you don't have to pay.
I'd trust Wordtracker over Google any day for accuracy on search term numbers, Google's numbers and also lateral searches are almost non-existent.
The danger in using Google CP is that you are trying to do SEO and using PPC results... the two things are very different in the way to approach.
| 5:01 pm on Sep 30, 2002 (gmt 0)|
They are both very poor. WT's flaw is that its counts are based on a mathmatical calculation of marketshare. The demographic makeup of Google isn't factored at all. in order for WT's numbers to be correct, you must buy into the notion that the type of people searching at dogpile and metacrawler are the same type of people that use Google. The problem is, that's not the case. Search behavior is not constant across different search properties.
Goole's flaw is that they simply do not show accurate numbers for searches. On a regular basis, I've found that keywords the show up in a client's log files frequently enough to be included in Google's keyword suggestion tool, don't ever make it in. The tool tells me no one searches for the phrase, yet we see hundreds of visitors per day for the phrase.
| 5:26 pm on Sep 30, 2002 (gmt 0)|
Thanks for the replies so far.
JayC: I had a look at past threads, as you advised, and I found some further thoughts, but unfortunately nothing I feel is conclusive (but then, perhaps nothing can be). Most of the past cases are ones in which the search query numbers on say, Wordtracker and Google are not identical but indicate roughly the same number of searches, and therefore, actually "agree" with one another. My situation is a bit different, since, in my case, what I'm getting is that my most searhed on keyword in Google according to Wordtracker is the least searched on in Google according to AdWords. So, in other words, in my case, the numbers don't agree at all; there is very a large discrepancy of about 1000 searches/day (30,000 searches/month or more than 50%, in this case). In this situation, would you suggest I trust the Wordtracker numbers or AdWords numbers? Note that I'm doing a technical SEO campaign here and not an AdWords PPC campaign, so the commitment to the keywords I choose is necessarily longer term; if I target the wrong keywords, I won't know it for a few weeks, at the least.
webdiversity: Thanks for the specific figures. They're quite helpful. As I note above, and as you note, I'm doing a technical SEO campaign, and trying to use AdWords figures for this purpose, rather than for an AdWords campaign. It seems that, in this situation, you're suggesting that I go with the Wordtracker figures over the AdWords figures.
Any further input with would be valuable.
| 6:00 pm on Sep 30, 2002 (gmt 0)|
Thanks for the reply. These are some very interesting observations. The WT assumption that searcher demographics (and therefore, query distribution) is the same across different engines, is indeed wrong. This is why I'm inclined to go with AdWords' numbers. My reservation with going this route, is that (just as you suggest) AdWords numbers are not very accurate or complete, based on my experience (and the fact that e.g. exact match and phrase match yield identical results).
All that said, and based on your experience, would you advise me to go with the numbers coming out of AdWords versus Wordtracker or vice versa? I wish I could target all of the keywords suggested by both AdWords and Wordtracker for Google, but that is practically impossible right now (and also has implicataions for site URL selection, etc.). Please advise.
| 5:50 am on Oct 1, 2002 (gmt 0)|
My advice would be to not focus so much on search count numbers in general.
While WT doesn't do a great job at predicting the actual number of searches that take place for a given term at a given engine, it does do a good job of ranking a group of terms in their basic order of popularity.
By that I mean that if you select a keyword set of 20 different phrases and you are able to get fairly equal distribution in your listings, a couple months down the road, your keyword referral report showing the top 20 phrases bringing visitors to your site will look very close to the list you pulled out of WT.
However, once you've established the positions for your original set, you will often find the real money terms in your log files. These are the phrases that neither tool told you about. You can only find them by tracking all the secondary matches your original optimization efforts produced.
So if you go into the project with the idea that that your original keyword research is just Phase 1, and the purpose of that research is to put you into the basic ballpark so you can find all the hidden gems that will really produce great results, then the whole process becomes a lot less stressful. :)
| 5:50 am on Oct 1, 2002 (gmt 0)|
In addition to WT's small sample size, and the demographic differences between WT and Google, there's something else to be considered... The Google AdWords Select tool, if that's what you're using, isn't reporting searches. It's reporting expected AdWords clickthroughs... and it's reporting rough approximations only.
The original AdWords CPM tool was reporting expected searches, but it was notoriously inaccurate. It was based on projections by Google that were admittedly (I have this from a Google rep) never updated.
You might be able to calculate expected searches with the Adwords Select tool data if you had additional information. You'd need to know what the AdWords clickthrough rate might be, in relationship to searches, for a well-crafted ad in the various positions. Perhaps veterans of the original AdWords program have some figures that could be applied to current stats. If the AdWords clickthrough rate ranged, say, between 3% to .5%, for example, multiplying the expected number of clickthroughs by 33 to 200 would give you an approximate number of searches.
>>Often the discrepancies pose big strategic road blocks. For instance, the keyword in my keyword set that Wordtracker indicates is generating the most searches via Google, AdWords indicates is generating nearly the least number of searches.<<
The discrepancies could well be a function of how satisfactory the search results are. I would assume that AdWords clickthroughs go up with search frequency, but not always. They might also go up when searched terms return unsatisfactory serps, and go down when the serps satisfy the searcher.
In areas where the search terms are ambiguous and might describe several different things, I've observed that the AdWords Select numbers are proportionately a lot higher than what I'd expect from the Overture numbers. I'm wondering whether anyone else has noticed this.
I've never assumed that any of these tools is useful for anything except indicating the probable relative number of searches, and even there you've got to be careful.
Further complications... the high threshold of the AdWords tool makes it useless for predicting niche searches. And the AdWords stats will become harder to relate to Google searches the more AdWords is used by other engines.
| 10:36 pm on Oct 1, 2002 (gmt 0)|
Thanks. First, a note: When I refer to AdWords keyword query counts, I am talking about the query counts coming out of the CPM AdWords control panel (not out of AdWords Select).
So, from the replies here, in sum, it would seem that the suggestion is to go with Wordtracker query numbers over AdWords numbers for Google SEO. Please let me know if I'm drawing the wrong conclusion here.
It would seem that Google's figures are out of date, less granular (exact matching, phrase matching, etc. yield the same numbers), and based on the collective experience of everyone who put down their thoughts, the truth is simply closer to what Wordtracker suggests, at least in terms of the relative distribution of searches, even if WT's exact numbers are not 100% precise (which they never could be expected to be anyway). Is that correct? Anyone object to this analysis?
| 5:55 am on Oct 3, 2002 (gmt 0)|
>>It would seem that... the truth is simply closer to what Wordtracker suggests.... Is that correct? Anyone object to this analysis?<<
Lorraine - I keep hoping someone else will jump in and answer your question. It seems to me that WebGuerrilla already has....
|While WT doesn't do a great job at predicting the actual number of searches that take place for a given term at a given engine, it does do a good job of ranking a group of terms in their basic order of popularity. |
I'm not happy with his answer, but I can't argue with it. I don't trust WordTracker's demographics or that tiny sample size... so I tend to use Overture and Google in combination, along with gut instinct, a lot of skepticism, parallel searches on similar terms, PPC data, etc. I may be forced to use WT more than I'd like to. The truth is that right now we really don't have a good targeting tool that lays it all out for us, especially for small niche targets.
Another problem with AdWords Select data, by the way, is that in addition to reflecting search popularity and how satisfied searchers are with the serps they see, it also reflects what advertisers choose to target.
Incidentally, where are you finding original AdWords CPM data?... or is this data you collected before the program was discontinued? For high volume phrases, it made some distinctions that I at least considered along with everything else, and it had been part of my mix of tools.
| 3:50 am on Oct 11, 2002 (gmt 0)|
Great thread! I think WebGuerrilla's advice is pretty spot on.
Do your initial research using a mixture of tools then start analysing your logs, reviewing your choices and even keeping an eye on what words your competitors are using. It all helps.
| 1:43 pm on Oct 16, 2002 (gmt 0)|
I typically use Overture numbers and multiply them by 4 to get estimated Google searches. It's arbitrary and that's why I like it. I then use WT to determine how many people are searching for plural vs. singular, and also to determine if more people are searching for "keyword1 keyword2" than "keyword2 keyword 1"