If you were to ask most SEO and PPC Marketers what tool they use to gauge how often a particular keyword is searched, the answer is probably the Google Keyword Tool. Originally only a part of AdWords, the Google Keyword Tool has been available to the public for keyword research for a number of years now, but most of the SEO and PPC folks I meet say the same thing about it: “It isn’t accurate”.
A discussion of the accuracy of Google’s keyword tool came up in a forum I was participating in recently, so today I thought it fitting to challenge that belief. In reality, the Google Keyword Tool is very accurate – the problem is most marketers aren’t doing the math properly.
Let’s look at this on a basic level first. You query a keyword and it tells you there are 18,000 local searches for that keyword, which would be about 600 searches per day. The complaint was that in spite of these numbers, the user – who was #1 for the keyword – was only receiving about 20 visitors per day. According to Google, 40% of searchers will click the first result so the user should be getting 240 searches per day, not 20, therefore the keyword tool is inaccurate, right?
Now lets look at this another way. Google returns 10 results, so even if we throw the 40% figure out the window and just divide the number of monthly searches by ten, that should still end up delivering about 60 visitors per day. Again, this “proves” the Google keyword tool is inaccurate because he’s receiving only a third of the expected traffic. Right?
The first mistake is in understanding how Google comes up with the 40% figure; this is an average across every keyword indexed in Google. What it really means is that Google’s algorithm is only accurately guessing what searchers want 40% of the time – the remaining 60% aren’t satisfied with the first result and are clicking one of the others. So the 40% rule isn’t a “rule” at all – it’s a network-wide average. On some phrases, it’s possible 90% are clicking the first result. On other phrases, less than 10% are clicking the top spot.
Now, lets look at the second mistake – forgetting to factor in paid results. For highly competitive terms, paid ads can take up the top 3 spots so even though you’re “#1” organically you’re actually #4 on the page.
Third, if your phrase has any local competitors, those local providers will be mixed in with the top search results as well, pushing your page down even further. And fourth, the additional PPC ads on the right sidebar offer even more choices. There can be 8 more of those on the right.
When you add it all up, for a highly competitive phrase there can be as many as 25 different choices for the user to click on. So 18,000 searches / 25 choices / 30 days = 24 searches per day. That’s pretty close to what the user was getting in practice, which makes the Google Keyword Tool pretty accurate if you ask me.
But that’s not the only factor to consider… an increasing number of searchers are signed in to their Google accounts while searching. This creates additional uncertainty, because if you’re signed in to Google then you’re searching Google via their secure site. Why is this an issue? Because the HTTPS protocol doesn’t pass referrer info to HTTP pages – making it impossible for your analytics to detect what keyword was used to bring visitors to you. You can try this yourself… do a search while signed in to Google and click through to one of your own sites, your site will not be able to detect what keyword brought the visitor to you. It will look like a direct type-in in your logs. So it’s also possible that many more people are landing on your page as a result of that keyword than you realize.
To get some more info on the subject, I started digging through the logs of one of my own sites. Coincidentally, one of the sites I manage is the perfect example that attests to what I’m saying here – a site with a domain that’s never going to be typed in, with no competition or local providers, and it’s such a low competition phrase that all ten of the page one results point to different pages on my site. According to Google, this keyword is searched 1,300 times per month. Since there are no PPC ads or local (map-based) results, and I own all ten of the spots, anyone who searches this phrase and clicks on a result will be landing on my site. I checked my stats for the previous month. If I combine the direct type-ins (“no keyword” traffic) and the known clicks from this keyword across all ten of the pages I’m ranking on page one for, I had 1,318 visitors as a result.
It’s not going to get much more accurate than that!
When you understand the variables and let go of expectations based on published averages, it turns out the Google Keyword Tool ends up being pretty accurate after all. The mistake people make is in “guessing” how many clicks they’ll get based on what the keyword shows as search volume. With anywhere between 10 and 25 destinations on page one of Google, and a large number of users still finding their way to pages 2 and higher as well, it’s not the keyword tool that’s inaccurate – it’s the method Internet marketers use to guess how many of those results will translate into clicks to their site that’s flawed.