Google’s search query reports provide PPC advertisers with two fantastic opportunities to improve the performance of their AdWords campaigns:

  1. Identify irrelevant keywords which can be added as negatives
  2. Identify new keyword opportunities for keyword expansion

The difficulty, however, is efficiently and reliably pulling out trends and insights from a raw search query report. According to Google, 25% of searches made each day are completely unique, and 70% of searches lie outside of Google’s Keyword Tool. While this suggests that the large majority of your search queries will have received only a handful of clicks (making trend-spotting extremely difficult), it also presents a great opportunity for identifying new keywords outside of the Keyword Tool.

This article will explore the techniques which can be used not only to identify negative keywords from a search query report, but also identify new opportunities for practical keyword expansion.

Not Enough Data

The main problem with a raw Google search query report is the sheer variety of searches. Each search query has minimal data, making it difficult to spot trends and insights. Suppose we had the following search queries amongst the mass of other search queries in your report:

  • holiday paris for winter 2011
  • holiday in winter 2011 to paris
  • paris winter 2011 package holiday

It is hard spot the opportunity ‘winter 2011’ amongst the noise of the search query report. It is hard to see the wood through the trees.

Something more is needed.

Search Query Aggregation

If instead, search queries were pulled apart into their individual words, and click data was then aggregated based on these individual words, we would be able to make more sense of search query reports. For example, suppose we could now identify the following search query themes:

  • holiday
  • paris
  • winter
  • 2011
  • package

We can now see that ‘winter’ and ‘2011’ are popular themes within our search query report. And by aggregating click and conversion data for these search query themes, we can quickly and easily identify the types of themes which are converting well, and those which are converting poorly.

Similarly, if the word ‘free’ was hidden amongst multiple search queries, it would have been difficult to realise that the word ‘free’ was a drain on your AdWords budget. But being able to look at the aggregate click cost and conversion data for all searches containing the word ‘free’, you could quickly and easily determine whether ‘free’ should be added as a negative keyword.

If you’re willing to hand over your search query data, Query Miner has developed a tool which allows you to see the words which are converting poorly, and should be added as negative keywords.

But search query analysis shouldn’t stop at identifying negative keywords.

Oh no.

The real power of search query is in identifying new keyword opportunities – the 70% of searches outside of Google’s Keyword Tool. Although looking at one-word phrases can be useful for identifying negative keywords, doing so would ignore other words in the user’s search query which may be crucial to determine the searcher’s intent.

For example, suppose you identified the following themes within your search query report:

Google AdWords Search Query ReportAggregating your click, cost, and conversion data for each of these themes would suggest that the words ‘cheap’, ‘discount’ and ‘weather’ are prime candidates for negative keywords, due to their poor conversion rates.

Similarly, such analysis might show that keywords containing the words ‘kids’ or ‘children’ are converting extremely well, and suggest that keywords containing the words ‘kids’ or ‘children’ should be expanded or have their bids increased.

But while this insight may be interesting, it is of little use in helping you expand your list of keywords. Just because searches which contained ‘kids’ or ‘children’ performed well for your existing keywords, does not mean they would perform well if you added new keyword variations of ‘kids’ and ‘children’.

Something more is needed.

Multiple Word Search Query Aggregation

Suppose that instead of analysing performance for each 1-word theme, you analysed performance for search queries containing 2-word phrases:

AdWords Search Query Analysis

Looking at longer phrases, you are now able can now better understand the searcher intent. You can now see that the performance of search queries containing ‘kids’ and ‘children’ is not universal among different searches, but instead depends on other words they are next to. When someone searches for ‘with kids’ or ‘with children’, conversion rate is extremely poor, but when someone searches for ‘without kids’ or ‘without children’, conversion rate is very high.

Similarly, with 1-word analysis, you may have jumped to the conclusion that the word ‘cheap’ is a poor-performer, and should therefore be added as a negative keyword. But by observing 2-word phrases which contain ‘cheap’, you can immediately see that not all ‘cheap’ search queries are poor-performers. Searches which contain the phrases ‘cheap vacation’ or ‘cheap holiday’ have in fact performed extremely well; it is only the search queries which contain the phrase ‘cheap flights’ which have converted poorly.

That’s all very interesting…

But while 2-word theme analysis might help you make more reliable observations about your search query themes, especially for negative keywords, it still is rather impractical for identifying tangible new keyword opportunities which can be added to your campaigns. Knowing that searches which contained the phrase ‘cheap vacation’ performed well, does not mean that adding the keyword ‘cheap vacation’ would deliver equally promising results. If you specialised in Paris holidays, you would get all sort of non-Paris visitors. You are once again ignoring the other crucial words in the search query which are essential for identifying the searcher’s intent.

Even carrying out a 3-word or 4-word search query theme analysis, you would no doubt uncover useless phrases such as ‘cheap holiday deals to’, ‘cost package for october’, and ‘in paris april 2012’, which would not be sensible to add as new keywords:

Finding New KeywordsWe need a way of highlighting new keyword opportunities which can quickly and efficiently be added to the account.

Something more is needed.


We need filters. Once we have identified our promising 2-word, 3-word, and 4-word phrases from our search query report, we need to filter them to ensure they contain essential qualifying words.

Exactly what words you will use as qualifiers will obviously depend on the specifics of your individual campaigns, but here I have filtered to only show phrases which contains the words ‘paris’ and ‘holiday’:

Google Keyword Research

I now have a list of sensible 4-word themes which I can add to my campaign as new phrase match keywords. And since each of these new keyword suggestions contain the words ‘paris’ and ‘holiday’, I can be confident that these keywords would deliver highly-targeted traffic.


There is a huge amount of data available to PPC advertisers via the search query report. Although we are seeing tools which help advertisers mine their search query reports for negative keywords, the real gold is in identifying new opportunities for keyword expansion. Being able to do so efficiently and reliably is surely the Holy Grail of search query analysis, and can be the difference between a mediocre and successful PPC campaign.

What techniques do you use to mine search query reports for new keyword opportunities? How do you ensure new keywords are qualified and relevant? Share your comments and experiences below.

Alan Mitchell

Alan Mitchell is a Google AdWords PPC specialist, based in Melbourne, Australia, with a proven track record at improving return on investment (ROI) from Google AdWords. Find out how Alan can help your business.

  • queryminer

    Hi Alan, love this article! It validates exactly what we are building/have built at queryminer.

    We agree that search query mining shouldn’t stop at negatives and we are very excited about using our algorithm for keyword expansion. I will try to explain below.

    Right now, queryminer is solving for multiple-word aggregation for poor performers. So, in your example queryminer would recommend “cheap flights” as the phrase match negative and not “cheap” because there are other good performing search queries with the word “cheap”.

    The next evolution for queryminer is keyword expansion. Taking our multiple-word pattern detection algorithm and flipping it on its head and suggesting new keyword opportunities which of course will include the qualifying-words as you have described.

    Thanks for the mention & link and your thoughts around search query mining.

  • Alan Mitchell

    @ queryminer

    Thanks for the clarification, and my apologies for overlooking Queryminer’s current capabilities for multiple-word negative keyword mining.

    I’m confident that with the huge amount of data now available to PPC advertisers, efficient analysis and mining techniques are now more necessary than ever, so look forward to Queryminer’s evolution into keyword expansion.


  • Joel

    Hi Alan
    Great article. I have a question – what tool do you use to pull out the occurrences of terms with one, two and three word strings? I have been looking for something that does this but do not know an effective tool/Excel method for doing this.
    For a few additional tips you can check out my approach to search query reports:

  • Alan Mitchell

    @ Joel

    I use my own Excel-based macros which I have developed over time to pull out individual and multiple-word themes, and aggregate click, cost and conversion data based on those themes.

    If you’re looking for an online tool to help you with your search query analysis, you should give Queryminer a try. They offer a free analysis of your search query report and will highlight the cost savings of any negative keywords they identify.


  • Attacat_Joel

    Another @ joel joining the conversation!

    Really good ideas there, but finding a decent way to group keywords in such a manner is the real sticking point for a lot of people (especially if you don’t want to give out your data). I was trying to bodge an Excel sheet recently but had to abandon for a while! @ Alan you should make yours available, might be some good linkbait there… 😀

  • Dmitry

    Hi, Alan!
    It’s really good article.
    But I have a question for you.
    According to my experience, the conversion depend not only on keywords, but also on ads and landing pages, which are tailored with appropriate keyword.
    Many times I observed, that the same keyword can give many conversions with one ad, and with another ad the same keyword convert badly.
    So it’s very important, with which ads your keywords are tailored.
    What do you think about this idea?

  • Alan Mitchell

    @ Dmitry

    I completely agree – it’s something I’ve written about before in my post on relevancy. While researching 1,000s of keywords is great, those keywords are not going to have the chance to achieve their full potential unless 1,000s of tailored ad messages are provided.

  • Enrique del valle

    Thanks for the info, the search query report is the key of any good PPC campaign.

    Alan, what do you think about search terms which receive no clicks? If we could know which keywords activate our ads with no clicks, it would be easy to significantly improve CTR and QS.

  • Alan Mitchell

    @ Enrique

    It would be great to see ALL searches which trigger your ads in the search query report, although I doubt Google would be willing to provide this information.

    Until then, we have to work with the data we have available, which I still think can provide an enormous amount of insight.