|Webtrends keyword clicks vs. KeywordDiscovery numbers|
how come they don't match?
| 8:17 pm on Sep 7, 2007 (gmt 0)|
In our WebTrends account, I was looking at keywords that brought traffic to a website over a 6-month period of time. Then, I did a search for the same keywords using the KeywordDiscovery tool. Not only were the number in KD tool much lower, but those low numbers were over a longer period of time (12-months) vs. the 6-month report I was viewing in WebTrends. Why would there be such a significant difference?
If KD tool pulls search data from 200+ search engines, and a majority of the keywords listed in WebTrends came from Google searches, how do I determine the overall popularity of these keywords during the time period I'm reporting? Shouldn't these numbers match up more closely?
| 8:37 pm on Sep 7, 2007 (gmt 0)|
|Shouldn't these numbers match up more closely? |
Not really, you're basically trying to compare apples and oranges: What one service says the number of keyword searches is supposed to be, and the keyword searches your logs show that actually drove visitors to your site.
Keyword tools are at best approximations that are best used for brainstorming, not baselining. They can help point out some weaknesses, say, if one shows a comparatively popular keyphrase for which your site should be receiving traffic and is not. They can also suggest some variations and related phrases you haven't thought of.
| 8:48 pm on Sep 7, 2007 (gmt 0)|
That makes sense, and I understand that. It just seems strange that the numbers in the keyword tool would be so low. If anything, I would have expected them to be inflated (more like Overture).
So, basically you are saying that I should throw away the KD tool search report if I am trying to prove the popularity of a specific topic, product or service over the last 6 months (based on keywords searched)? Or would you show both, since the KD tool shows additional keywords searched but maybe did not send traffic to this site?
The KD tool does show some nice graphs of trends (what months people searched the most), but it seems that this would only confuse the corporate folks who will be reviewing this report. Any other thoughts on the best way to show this?
| 1:47 am on Sep 8, 2007 (gmt 0)|
The KD tool may be showing accurate trend patterns and may be giving good comparisons from one keyword to another (keyword A gets searched 3 times as often as keyword B). It MAY be accurate as I said. KD says it gets data from 180 search engines but if you read the small print it says it gets logs from some of them and the others are from "panel data." My guess is that it gets logs from very very small search engines and uses some kind of panel data, examining searches done by people on the panel on the big search engines. That way, it can say it's getting data "from" searches on Google etc. Then, it uses some kinds of multipliers based on demographic data to multiply its data into estimates for the entire planet.
If you think most of your visitors are likely to be using Yahoo and Google then you should stick with their own keyword suggestion tools for whatever country your visitors are probably in. If on the other hand your visitors tend to use obscure search engines you might get better keyword suggestions from something like KD.
If you really want to use KD to build a business case for various keywords, then you can try to pro-rate their data. If there's a keyword that's already bringing in traffic, and that you know is already in the top ten on Google etc (i.e. visible to all who search on that term), and if you think the text of your listing is fairly compelling, then compare the number of visits reported by WebTrends for that word to the number of searches reported by KD. Divide one by the other to get a multiplier. Apply that multiplier to stats for other keywords that KD has, to get an idea of the possible number of site visits you COULD get from the other keywords (assuming a top ten position and reasonably compelling listing text).
Remember that a lot of people repeat searches using variations, so related keywords will cannibalize each other.