How Accurate is the Adwords Tool for Monthly Searches Estimation?

[akasher]

Let's say, for example, that google adwords estimates 3k searches per month for an exact match keyword.

If I ranked #1 in google for that given keyword, (and supposedly yahoo and bing as well), would I expect to receive AT LEAST 3k uniques per month?

Or is the number higher because of the yahoo and bing searches as well?

[dburk]

Hi akasher,

No, search volume is before click-throughs. On average only about 23% of searches will click on the #1 result. That's just an average and it will vary for different types of keywords. Generic type keywords typically have a slightly lower CTR.

[akasher]

Originally Posted by dburk

Hi akasher,

No, search volume is before click-throughs. On average only about 23% of searches will click on the #1 result. That's just an average and it will vary for different types of keywords. Generic type keywords typically have a slightly lower CTR.

Thanks. I am actually ranking for very specific, mini niche keywords, so I think I"ll have a decent CTR.

[TryBPO]

23% Seems a bit low from that data we've both seen and read...approximately 40% is closer to what we've found. Either way, he's right in that it will be a percentage of that exact match search volume you found using the GKT and that percentage is less depending on how far down the page you're ranked.

The GKT is a pretty good rough estimate on the number of searches. Make sure to see if most of the traffic came in one month, is seasonal, etc. as well...that might give you a good idea as to whether the numbers are skewed due to a market blitz, keyword being in the news, etc.

[dburk]

Originally Posted by TryBPO

23% Seems a bit low from that data we've both seen and read...approximately 40% is closer to what we've found.

Hi TryBPO,

I have seen certain types of keyword searches the can generate a 40% CTR. For example a search for a specific domain name like facebook will generate a relatively high CTR for the facebook.com website. However, when someone searches for a product or service type of kewyord the CTR is rarely over 20% on average.

I have never seen any empirical data that supports the 40% claim. I, like you, have seen plenty of folks on this forum repeat the 40% number, but so far no one has cited a source to support it. When you look at the largest publicly available data-sets they generally range from 17%-23% CTR for position #1.

If you have seen a credible source to support that 40% claim I'd love to see the source. Can you share?

[gman0156]

I think if you site has many interesting page. your pageviews will be much more than no.of visitor.
for example, if you can get 1,000 visitor a month with avg 3 pageview/visitor so you can get 3,000 pageviews.

[dburk]

Originally Posted by gman0156

I think if you site has many interesting page. your pageviews will be much more than no.of visitor.
for example, if you can get 1,000 visitor a month with avg 3 pageview/visitor so you can get 3,000 pageviews.

And how is that information relevant to this thread?

[Joelrunner]

I'd like to also point out that all of the keyword estimation tools (including Google's own tool) are wildly inaccurate.

I did a webinar with Perry Marshall and we demonstrated how they can be off (in either direction) by as much as 50%.

In essence, if it's going to take a large SEO effort to rank for any given term, it would be well worth your while to simply buy some exact-match traffic to verify that you can actually monetize visitors. It's never as expensive as you think - and if you don't convert that traffic, you'll be glad you didn't spend dozens or hundreds of hours optimizing for that term.

[chimehost]

If google is showing 50,000 exact searches, you should take that number and divide it by 2 or 3. Use that number and times it by 30-40% and that should be the amount of user who will visit your site if you ranking no 1. After that you have take into account your own website conversion rate.

[dburk]

Originally Posted by chimehost

If google is showing 50,000 exact searches, you should take that number and divide it by 2 or 3. Use that number and times it by 30-40% and that should be the amount of user who will visit your site if you ranking no 1. After that you have take into account your own website conversion rate.

Hi chimehost,

I disagree, I think that would be extremely overestimating the search traffic from most keywords. Any estimate over 25% would be overly optimistic, except for certain types of searches, like a search for a specific company's website. This opinion is based on both observational study as well as my own experiments. What is the basis for your assertion?

Another consideration is traffic from other keywords, which I am excluding in my estimates for traffic from a specific keyword. We might extrapolate traffic from alternate keywords, but that is separate calculation with many variables.

[mori22]

market samurai has a feature called SEOT that estimate the traffic under a first place assumption

[chimehost]

Originally Posted by dburk

Hi chimehost,

I disagree, I think that would be extremely overestimating the search traffic from most keywords. Any estimate over 25% would be overly optimistic, except for certain types of searches, like a search for a specific company's website. This opinion is based on both observational study as well as my own experiments. What is the basis for your assertion?

Another consideration is traffic from other keywords, which I am excluding in my estimates for traffic from a specific keyword. We might extrapolate traffic from alternate keywords, but that is separate calculation with many variables.

I have see CTR anywhere from 10%-55% for Google search. You have to keep in mind if your title / meta description is not interesting, users might skip your site. Most people make their Title and Description optimized for seo but never for the users. When users search in Google, sometimes they all look the same. It's hard to come with just a % but this is just to get an idea before users start focused on that keyword and investing their hard earned money.

[TryBPO]

Originally Posted by dburk

Hi TryBPO,

I have seen certain types of keyword searches the can generate a 40% CTR. For example a search for a specific domain name like facebook will generate a relatively high CTR for the facebook.com website. However, when someone searches for a product or service type of kewyord the CTR is rarely over 20% on average.

I have never seen any empirical data that supports the 40% claim. I, like you, have seen plenty of folks on this forum repeat the 40% number, but so far no one has cited a source to support it. When you look at the largest publicly available data-sets they generally range from 17%-23% CTR for position #1.

If you have seen a credible source to support that 40% claim I'd love to see the source. Can you share?

I totally agree with you that a search for "Facebook" will get an extremely high percentage of searchers who click on the top result. (Maybe not even BECAUSE it's the top result...if facebook.com happened to rank #2 but was the MAIN FB site, I think the #2 spot would get a ton of clicks)

Here's some supporting information:

SEObook mentioning leaked AOL data (It's not Google, but probably a large enough data set to make a fair comparison?)
Google Keyword Click Data by Search Ranking Position

Cornell eye tracking study on SERP's:
http://www.cs.cornell.edu/People/tj/...a_etal_04a.pdf

Others mentioning and graphically representing the eye tracking study (less trusted sources):
» Distribution of Clicks on Google’s SERPs
The Distribution of Clicks in Google's SERPs

More data:
How your Google ranking affects your Click-Through Rate (CTRs)? | {SEO Company} SEO | SEM | Search Engine Marketing & Optimisation | Web Design | iClick Media

Interesting mention on how +1 might effect user behavior for CTR on the SERPs (trusted source):
Google +1 Increases CTR in Organic SERPs, Ranking Boost Likely [Study]* - Search Engine Watch (#SEW)

[Tuffy22]

I think the exact search estimates are fairly reliable, but figure a 1st ranking is going to result in anywhere between 15-45% CTR. Several variable come into play such as how narrow the niche is, as well as an inviting Title and Meta Description.

[Fraggler]

Originally Posted by dburk

Hi TryBPO,

I have seen certain types of keyword searches the can generate a 40% CTR. For example a search for a specific domain name like facebook will generate a relatively high CTR for the facebook.com website. However, when someone searches for a product or service type of kewyord the CTR is rarely over 20% on average.

I have never seen any empirical data that supports the 40% claim. I, like you, have seen plenty of folks on this forum repeat the 40% number, but so far no one has cited a source to support it. When you look at the largest publicly available data-sets they generally range from 17%-23% CTR for position #1.

If you have seen a credible source to support that 40% claim I'd love to see the source. Can you share?

Here is some screenshots from my Webmaster account for a site in the weight loss niche:

http://dl.dropbox.com/u/5199729/ctr1.png
http://dl.dropbox.com/u/5199729/ctr2.png

They aren't huge search numbers but gives your a rough idea. The CTR is definetly higher than 17-23% for a number 1 spot. If you want more data I'll happily export all 3300+ keywords (minus the keywords) for you have a look at.

I have also found that Google's search data closely matches what I end up when I'm on the front page with my Webmaster Tools numbers.

[Fraggler]

I added a spreadsheet with the numbers from the screenshot above.

http://dl.dropbox.com/u/5199729/webm...chqueries.xlsx

Back to Adwords Tool vs Actual Numbers:

Here are some examples from my list to give some small indication.

GWT: 12,000. AWT (Global): 12,100
GWT: 12,000. AWT (Global): 6,600
GWT: 60,000. AWT (Global): 60,500
GWT: 50,000. AWT (Global): 27,100
GWT: 4,500. AWT (Global): 5,400
GWT: 5,500. AWT (Global): 3,600
GWT: 1,600. AWT (Global): 1,900

As you can see, there is some variation but it swings both ways. It isn't as inaccurate as some make out though.

[dburk]

Originally Posted by TryBPO

I totally agree with you that a search for "Facebook" will get an extremely high percentage of searchers who click on the top result. (Maybe not even BECAUSE it's the top result...if facebook.com happened to rank #2 but was the MAIN FB site, I think the #2 spot would get a ton of clicks)

Here's some supporting information:

SEObook mentioning leaked AOL data (It's not Google, but probably a large enough data set to make a fair comparison?)
Google Keyword Click Data by Search Ranking Position

Hi TryBPO,

Thanks for posting this data that supports my previous post.

The AOL data has just under 23% CTR for position #1 in search. And SEO Book points out, just as I did, that you can often expect higher CTR when search is for a brand name website. The implication is that non branded website searches, on average, are low enough to bring down the overall average to lower than 23%.

So if you are doing keyword research for a product category or niche, you can expect substantially lower than 23% CTR on average.

Originally Posted by TryBPO

Cornell eye tracking study on SERP's:
http://www.cs.cornell.edu/People/tj/...a_etal_04a.pdf

There is no click through rates mentioned in this document and the searches were not real-world searches, but controlled laboratory searches with the specific keywords dictated during the experiment, 50% of which were branded website searches. There is no way to even etrapolate realistic CTR data from this study, probably why it's not mentioned in the white paper.

Others mentioning and graphically representing the eye tracking study (less trusted sources):
» Distribution of Clicks on Google’s SERPs
The Distribution of Clicks in Google's SERPs

Again, no CTR data here.

More data:
How your Google ranking affects your Click-Through Rate (CTRs)? | {SEO Company} SEO | SEM | Search Engine Marketing & Optimisation | Web Design | iClick Media

This appears to be a mislabeling of data. I looks like they may have labeled % of clicks as CTR, those are two very different things. The source data was put behind an opt in form so I can't be sure. Did you download the data? And did it include the search volume?

Originally Posted by TryBPO

Interesting mention on how +1 might effect user behavior for CTR on the SERPs (trusted source):
Google +1 Increases CTR in Organic SERPs, Ranking Boost Likely [Study]* - Search Engine Watch (#SEW)

Interesting, but no data that is relevant to this situation.

So, I guess you weren't able to find any data that supports the 40% claim. Neither could I.

[Fraggler]

Originally Posted by dburk

I looks like they may have labeled % of clicks as CTR, those are two very different things.

Just so you know, GWT labels them as % of clicks as well. Feel free to calculate the CTR from that spreadsheet I uploaded, you'll find they are very close to the same.

Actually: the breakdown of the top 10 spots does display the CTR.

40% is high but I really think < 23% is very conservative.

[TryBPO]

Originally Posted by dburk

Hi TryBPO,

Thanks for posting this data that supports my previous post.

The AOL data has just under 23% CTR for position #1 in search. And SEO Book points out, just as I did, that you can often expect higher CTR when search is for a brand name website. The implication is that non branded website searches, on average, are low enough to bring down the overall average to lower than 23%.

So if you are doing keyword research for a product category or niche, you can expect substantially lower than 23% CTR on average.

Very confused on how you came to that conclusion. The leaked AOL information shows 42%+ going to the #1 natural/organic position. The comments do mention the 42%+ being before local, universal search, etc.

Check this out for the specific numbers:
Jim Boykin's Link Building Blog | Click Rate for Top 10 Search Results

What I didn't see pointed out in the articles is whether this was counting the FIRST or ALL clicks from the SERP. What about those who click the first result, don't find what they're looking for, click the second result, etc.

Originally Posted by dburk

There is no click through rates mentioned in this document and the searches were not real-world searches, but controlled laboratory searches with the specific keywords dictated during the experiment, 50% of which were branded website searches. There is no way to even etrapolate realistic CTR data from this study, probably why it's not mentioned in the white paper.

If you look at the graph, you'll see that the #1 result was selected 3 times more often than the #2 result. I didn't see the actual numbers, but from the graph it looks pretty close to 55% and is tracking even higher than the AOL data. (Actually, this graph represents the clicks: » Distribution of Clicks on Google’s SERPs)

Originally Posted by dburk

This appears to be a mislabeling of data. I looks like they may have labeled % of clicks as CTR, those are two very different things. The source data was put behind an opt in form so I can't be sure. Did you download the data? And did it include the search volume?

I didn't opt-in and didn't see an opt-in there either, but if you're looking for more information about the Optify study, check this out:

How Much Money is a Top Google Ranking Worth to Your Business?

Optify Study: 36%
Cornell Study: 56%
AOL Data: 42%

It also discusses brand bias and that a full 84.2% us those searching for a brand will click the first position. It goes on to show, though, that Jakob Nielson's research showed that even when the top two listings are switched, it goes from 42%/8% to 34%/12%...pointing to a bias to the first position and not necessarily having it mostly be about the brand:
The Power of Default Values (Jakob Nielsen's Alertbox)

Originally Posted by dburk

So, I guess you weren't able to find any data that supports the 40% claim. Neither could I.

I don't have any stake in this...it's not an issue to me whether you ignore the research or not...but claiming I didn't present reliable data is a bit of a reach. I think the Jakob Nielson study is particularly relevant...let me know what you think.

[dburk]

Originally Posted by TryBPO

Very confused on how you came to that conclusion. The leaked AOL information shows 42%+ going to the #1 natural/organic position. The comments do mention the 42%+ being before local, universal search, etc.

Check this out for the specific numbers:
Jim Boykin's Link Building Blog | Click Rate for Top 10 Search Results

What I didn't see pointed out in the articles is whether this was counting the FIRST or ALL clicks from the SERP. What about those who click the first result, don't find what they're looking for, click the second result, etc.



If you look at the graph, you'll see that the #1 result was selected 3 times more often than the #2 result. I didn't see the actual numbers, but from the graph it looks pretty close to 55% and is tracking even higher than the AOL data. (Actually, this graph represents the clicks: » Distribution of Clicks on Google’s SERPs)



I didn't opt-in and didn't see an opt-in there either, but if you're looking for more information about the Optify study, check this out:

How Much Money is a Top Google Ranking Worth to Your Business?

Optify Study: 36%
Cornell Study: 56%
AOL Data: 42%

It also discusses brand bias and that a full 84.2% us those searching for a brand will click the first position. It goes on to show, though, that Jakob Nielson's research showed that even when the top two listings are switched, it goes from 42%/8% to 34%/12%...pointing to a bias to the first position and not necessarily having it mostly be about the brand:
The Power of Default Values (Jakob Nielsen's Alertbox)



I don't have any stake in this...it's not an issue to me whether you ignore the research or not...but claiming I didn't present reliable data is a bit of a reach. I think the Jakob Nielson study is particularly relevant...let me know what you think.

Hi Justin & Joe,

Thanks for clarifying where you got the 40% number from.

Please allow me to point out the error in your calculation. You have used the percentage of clicks rather than the click through rate. Those are two very different numbers. It is a common error so please don't feel bad.

The click through rate is calculated by dividing the number of impressions by the number of clicks.

For example in the AOL data you cited at Jim Boykin's blog you can see that the number of clicks for position #1 was 2,075,765 and the number of impressions was 9,038,794.


2 075 765 / 9 038 794 = 0.229650659

We seem to be in total agreement on the brand bias generating much higher CTRs. That is why I suggested that non branded searches must be lower than the overall average CTR of ~23%.

@Fraggler,

We were specifically talking about averages from empirical data, not top performing keyword data.

[Fraggler]

Originally Posted by dburk

We were specifically talking about averages from empirical data, not top performing keyword data.

The spreadsheet has the data for an entire site, a real site; but because I didn't do an experiement the data isn't valid? The only other data I have ever seen is the AOL data, which you correctly point out shows the 23% CTR for a #1 spot.

How old is that data? Did the AOL engine back then have the same interface as Google does now? Are Google searchers different to AOL searchers?

Nothing is removed from the file I provided except for the keyword names, i.e. it is not top performing keyword data. You can see keywords with > 1000 impressions and ones with < 10. It is actually full of duds.

My data might not follow scientific process but I know it does show that there are generic keywords getting a lot more than 23% of total search impressions in the weight loss niche.

My file was nothing but an example showing that I experience greater than 23% of total search impressions for generic keywords. Maybe others would like to share too?

[TryBPO]

What does hold true, though, is that looking at a comparison between the number of clicks in one position as compared to another gives you a comparison...but that comparison doesn't hold true when you're comparing to the number of searches performed.

You cleared up something I didn't think I needed clearing up! lol

@Fraggler - I do wonder whether Google's data is similar...interesting to see the heatmap showing a much tighter view for Google in the top lefthand corner of the page...I wonder what effect that has.

[dburk]

Originally Posted by Fraggler

The spreadsheet has the data for an entire site, a real site; but because I didn't do an experiement the data isn't valid? The only other data I have ever seen is the AOL data, which you correctly point out shows the 23% CTR for a #1 spot.

How old is that data? Did the AOL engine back then have the same interface as Google does now? Are Google searchers different to AOL searchers?

Nothing is removed from the file I provided except for the keyword names, i.e. it is not top performing keyword data. You can see keywords with > 1000 impressions and ones with < 10. It is actually full of duds.

My data might not follow scientific process but I know it does show that there are generic keywords getting a lot more than 23% of total search impressions in the weight loss niche.

My file was nothing but an example showing that I experience greater than 23% of total search impressions for generic keywords. Maybe others would like to share too?

Hi Fraggler,

Your data is useful for analyzing your particular site, however those numbers are not comparative to the data that we might expect from a broader based sampling. Your data is purely anecdotal.

Additionally, you may notice that the data reported from Webmaster Tools does not include impressions where no clicks were generated, it is not a true representation of CTR based on total search volume, instead it is based on searches that resulted in a click through either to your website, or another. In presenting it this way Google is able to mask the true CTRs, which seems to be a metric that Google wants to keep private.

Another consideration when viewing your data is that we cannot see how much of your results are from branded vs. generic searches. For the purpose of the OP's question, branded search data should be excluded to get a truer representitive number for calculating expected traffic from a particular generic (non-branded) keyword.

While the AOL data is dated and cannot be relied upon to represent current search results, it is based on search results that were powered by Google's search engine, and no comparable sampling has ever been made public, before or since. I am fairly certain that subsequent features like Google Autocomplete and Google Instant have had a significant influence on CTRs, be it positive or negative, as well as improvements in the search algorithm.

While search algorithm changes have likely increased CTRs for certain keywords, particularly branded keywords, I don't think they have changed that much for generic keywords. It seems to be an issue influenced primarily by human nature, and I just don't see human nature changing all that much. If anything, I see branded searches becoming more common behavior in users and this has no real influence on the calculation one might use for evaluating the traffic potential of a generic keyword.

[MikeFriedman]

I think the Google Keyword Tool should come with the same disclosure required on investments.

Past performance does not guarantee future results.

The GKT is presenting those keyword search numbers largely on what has happened in the past. That could change drastically in the future.

[PartsDr]

The adwords keyword tool is not very accurate, but it will give you somewhat of an idea. I would give it a plus or minus 50% accuracy. I do find it a very useful tool though.

[LemonSqueezy]

Honestly, it isn't all the accurate but you also have to realize that traffic fluctuates between months, weeks, etc. I use it more for competition than anything else, and to get a general sense of how many visitors could come through

[TryBPO]

Originally Posted by MikeFriedman

I think the Google Keyword Tool should come with the same disclosure required on investments.

Past performance does not guarantee future results.

The GKT is presenting those keyword search numbers largely on what has happened in the past. That could change drastically in the future.

That would make sense if the GKT were built to help those looking to determine their potential traffic and revenue, but the GKT is built to help AdWords advertisers determine good keywords and costs to use for their AdWords campaigns. I guess your point still holds true for them, though.

[dburk]

Originally Posted by Joelrunner

I'd like to also point out that all of the keyword estimation tools (including Google's own tool) are wildly inaccurate.

I did a webinar with Perry Marshall and we demonstrated how they can be off (in either direction) by as much as 50%.

In essence, if it's going to take a large SEO effort to rank for any given term, it would be well worth your while to simply buy some exact-match traffic to verify that you can actually monetize visitors. It's never as expensive as you think - and if you don't convert that traffic, you'll be glad you didn't spend dozens or hundreds of hours optimizing for that term.

Originally Posted by PartsDr

The adwords keyword tool is not very accurate, but it will give you somewhat of an idea. I would give it a plus or minus 50% accuracy. I do find it a very useful tool though.

Hi Joelrunner & PartsDr,

I have to disagree with these opinions. I believe the GAKT is highly accurate, after all, it is reporting actual data, not estimates. I don't think it is accurate on your part to assert it to be an inaccurate tool, perhaps inaccurately applied?

As MikeFriedman pointed out you cannot rely on past data to accurately predict future performance, you must take into consideration short and long term trends, seasonality, unusual or unexpected events, etc.

Furthermore, it seems clear to me now that many folks using that tool do not understand or allow for realistic CTRs that take into account the type of keyword.