Keyword Research Guesswork With Legacy, Adwords and API Tools

Yes, you read the title of this post correctly. Keyword research data accuracy can be extremely difficult to gauge, even with access to the best keyword research tools. Today, I’ve been testing the output of our (internal) Google Adwords API tool.

Having the tool allows much faster automation of large keyword research projects for Builtvisible clients, but I’m still interested in the differences between the API data, the public legacy Google Adwords tool data and the new Google Keyword Tool data.

2009 Legacy Vs Beta Shootout

Last year we compared the legacy Google Keyword Tool to the newer, beta version. Back then there were some significant differences in the data and things haven’t changed all that much. In this test there’s such variation in the numbers that there’s no point trying to spin fancy correlative measures or any statistics whatsoever in this post. The charts speak for themselves, and you can take a look a the data by downloading it here.

The only important thing to remember is both the external tools give local and global search volume from the previous month. The API data provided in this spreadsheet is a monthly average (over 12 months) for the UK locally and globally.

pivot table

Google UK Local Search Volume

Google local search volume

Click to enlarge the chart – the difference in search volume is amazing.

Google Global Search Volume

Google Global Search Volume

Again, click to enlarge the chart – differences in search volume numbers are significant.

Which Tool is Best?

I’ve been feeling generally better about the reliability of the data from the API tool over the external tools. Much of that sense is based on an understanding that the data is averaged across the year – you’re far less likely to get tricked into making a decision based on a single, seasonally impacted value. Of course, the newer Adwords Keyword Tool allows you to export the seasonal (monthly) values too and in reality there probably is no “best” tool or approach. Good keyword research should take as much data into account as possible, including additional insights from Webmaster Tools and your clients paid search campaigns.

Keyword research is about common sense decision making with an awareness of the strengths and weaknesses of the data from your tools. Provided you can keep those strengths and weaknesses in mind and stay objective, you’re better off having the numbers than not at all. I think the topic of keyword research in SEO is so subjective that many SEM’s bring a difference of opinion on methodology.

Image credits:
Håkan Dahlström

Learn More

Builtvisible are a team of specialists who love search, SEO and creating content marketing that communicates ideas and builds brands.

To learn more about how we can help you, take a look at the services we offer.

Stay Updated

Follow: | | |

Tags: , | Categories: Research

11 thoughts on “Keyword Research Guesswork With Legacy, Adwords and API Tools

  1. Branko says:

    Nice insight Richard.

    First of all kudos for making all the data available :)

    Secondly, I have personally found the Google Search based Keyword tool to be the most accurate when it comes to the traffic volume estimates. All the Adwords-based tools take into account the numbers that stem from ad exposure on partner networks, so they are usually an overestimate of what happens in organic search. I have seen cases where the WMT impression data agreed with the Omniture data and those numbers were 10 times lower than what was reported in Adwords Keyword tools, while the search based keyword tool gave estimates which were more or less in the same ballpark area as the real impression data from WMT and Omniture.

    Since different niches get different amounts of exposure due to differing levels of profitablity, it is almost impossible to come up with anything close to a constant offset by which these tools over- or underestimate the actual search traffic.

    The big drawback with the search-based tool is that it will not show all the keywords, especially if they are single-word terms (like poker or casino).

  2. Hey Branko, nice to hear from you!

    Yes – I actually hope that, by making the data available someone might be able to find the time to add the search based keyword tool data, and possibly webmaster tools data into the mix. Of course that would require someone with the right data.

    I’d also have loved to have the time to document my thoughts on why these tools are all so different. I hope that, through comments and conversation we’ll come across something interesting.

    Speak soon!

  3. Hi Richard,

    Nice right up. The differences between the tools is quiet big even taking into account the closer results from the API. I think that no one is going to give away the exact figures as that would make finding the pot of gold at the end of the rainbow relatively straight forward. I really think you need to use a combination of tools from different sources and then as you suggested common sense comes into it. I have particularly liked the Google Webmaster tools data and combining that with other sources. I also think having a good understanding of the web sites domain are critical as this can play a big part in your interpretation of the data you get back.

  4. Tom says:

    Using a combination of seasonal trend data and common sense, it is possible to worm out the vast majority of the anomalies from these data pulls. An approach which can help make the process more beneficial is to consider each KW on its relative merits, thereby circumventing the need to scrutinize specific numbers.

    Although this means that you cannot report on specific potential revenue estimates and (to an extent) traffic targets, this does allow you to bucket terms into suitable ‘head’, ‘middle’ and ‘tail’ terms. It also means that even if the numbers are 10 times what they should be, their relative importance remains unchanged.

  5. Thanks for this timely (as always!) post, Richard. I use a combination of tools including Wordtracker, Keywordspy, Spyfu and of course adwords to compliment the data gathered from WMT and Analytics.
    I have to admit I am still most reliant on the external adwords tool because I “trust” it more simply because it is Google’s data.
    I’m not saying the data is accurate, but I would argue it is relevant when it comes to prioritising one keyword over another and the problem I have with all the other tools is that their data sets are smaller i.e. you get a lot more “not enough data” feedback.

    What I like about SB Keyword Tool is it tends to filter out the crap, but when trying to get keywords for SEO then I’m worried about missing a trick.

    However, when all is said and done, I shall take on Branko’s advice when looking at volumes.

  6. Here’s something to add to the mix…. I just did a search on Wordtracker to help me with keywords around “shirts” and the number one suggestion with the highest volume?????

    Short breaks!

    Either Wordtracker is so finely tuned to me and my location and has figured Uk, they like to take their shirts of when on holiday so “short breaks”


    Wordtracker has made some sort or association with the fact most of my clients are travel related


    …. answers on a postcode

  7. Haha – I’ll pop down to the post office now…

  8. Phil says:

    Hi Richard,

    Thanks for sharing the insights.

    You said:

    “The only important thing to remember is both the external tools give local and global search volume from the previous month. The API data provided in this spreadsheet is a monthly average (over 12 months) for the UK locally and globally.”

    Wonder if you could clarify something for me. Does the API return exactly the same data as the the external tools or is it pulling back something different? I’m not sure if you calculated the API data as an 12-month average in Excel or if the API is doing this i.e. averaging the data out?

    Reson for asking, is obviously it makes better sense to make decisions based on a 12-month average vs. the previous month.


  9. Hi Phil,

    The average is (to the best of my knowledge) how it comes back from the API by default. So – no calculations in Excel!

    We’re doing some work to improve our data collection via the API, driven in part by the impending death of the search based keyword tool.

    Hope that clarifies – and if you have anything you’d like to compare notes on – give me a shout.

  10. Bill Edwards says:

    Hi Phil,

    Thanks for this informative post. I’ve been booking up on Google’s history of publishing search volume data, and I can’t help but believe that they intentionally make this data inaccurate because they don’t want advertisers to stop bidding on keywords based on search volume. Do you know what Google’s reasoning is behind hiding this data? Is there a technical issue underneath the covers?

  11. matt says:

    Using a combination of seasonal trend data and common sense, it is possible to worm out the vast majority of the anomalies from these data pulls.

Leave a Reply

Your email address will not be published. Required fields are marked *