Google Trends reveals search engine psychographics

/

Since Google makes news faster than other Internet companies, almost everyone already knows about Google Trends, announced two days ago. In case you’ve been busy doing real work over the past two days: it enables you to type in up to five keyword phrases and shows you the relative number of searches done for those phrases on Google (but there are no numbers, so it is a graph without coordinates.) They also show major news events for those terms and do bar graphs for various countries.

Here was my first graph:

I was very surprised to see these results, because on Trellian’s Keyword Discovery, which captures a full year’s worth of data but only samples about 1/11th of the Internet, I got these results, showing “embedded linux” to be the larger term:

It’s possible, I thought, that Google is picking up all searches that included the term (as if they were doing what Google AdWords calls broad match), because KeyWordDiscovery shows many more derivative terms containing cygwin. However, that guess was wrong wrt Google Trends: if you want a broad match, you have to separate your terms with a pipe like this: (embedded linux)|cygwin.

The real issue seems to be about pitfalls that a service like KeywordDiscovery faces when they sample, and the skew of various kinds of words. KD’s FAQs say that they do use Google as part of their sample, but they also use many other engines. This probably works well for them for most situations except for the one I happend to try. After all, if your company is involved with embedded linux, or you want to understand the market or the competition, you may easily be a manager who could use Google or Yahoo or MSN, typing in embedded linux. On the other hand, my limited experience with real geeks (the kinds who would type in cygwin) is that they use Google almost exclusively.

Robbin Steif
LunaMetrics

Our owner and CEO, Robbin Steif, started LunaMetrics twelve years ago. She is a graduate of Harvard College and the Harvard Business School, and has served on the Board of Directors for the Digital Analytics Association. Robbin is a winner of a BusinessWomen First award, as well as a recent Diamond Award for business leadership. You should read her letter before you decide to work with us.

  • Anonymous

    Heh. Real geeks know what cygwin is already and don’t need to type it in, in any search engine.

    😉

    Silly jokes aside, I suspect the results would lie more with the type of searches being conducted. “embedded linux” is more likely to be a standalone phrase.
    “cygwin” is, IMHO, more likely to be part of a more detailed question. Using google to index into bugzilla reports and the like. Does your analysis on Trellian’s KD shed more insight into that facet?

    The other part of possible relevance in the sample you’ve shown, is the country breakdown. That’s a fascinating little exercise. Keeping in mind that the relative graph lengths are normalised, I feel that this is actually a more useful display than “raw numbers”.

    Partly I suspect the problem may also lie in the sampling methodologies. TKD, presumably, has sampling stations in prominant ISPs and the like – similar to HitWise et al.
    Google has access to the lot.
    So Google’s reduced sample set can be more accurate, TKD’s may be unable to capture queries from large corporates or Govt – the very ones who are more likely to have geeks looking for “cygwin” as part of their support role.

    When we had a HitWise subscription it was quite obvious that this sampling bias was happening. Based on my direct knowledge of various networks and from my analysis of our various systems.

    But trends is an amazing tool. It’s quite clear to me why Google would release such a thing: it sets up an amazingly positive feedback loop between Google, people like us, and our site visitors.

    The full implications are still sinking in, but I and my colleages at work have just been handed the most amazing piece of *free* information to totally revolutionise how we deliver to our target audiences.

    And that rocks big time! 🙂

    Steve

  • http://www.lunametrics.com/blog LunaMetrics Blog

    “cygwin” is, IMHO, more likely to be part of a more detailed question.

    You don’t mean, it’s more like to to be part of a longer phrase, right? I started out with that assumption but eventually understood that Trends, like KD, was only showing me the information for people who had typed in that one word. To say the same thing in a different way, it’s the equivalent of a Google AdWords exact match, like this: [cygwin]. So it is really an apples to apples comparison. I really do believe (although “believe” is the operative word here) that the skew is in the KD sampling.

    As for countries, it is interesting. But I am so tired, and have miles to go before I sleep. Another day.

    Robbin

  • Anonymous

    You don’t mean, it’s more like to to be part of a longer phrase, right?

    Yes I did. Poor word choice. 🙂

    That’s a useful insight Robbin: that it’s only showing exact words. I had not made that distinction. Thanks!

    As you say, another day. I suspect there is much yet to learn with this tool.

    Steve

Contact Us.

LunaMetrics

24 S. 18th Street, Suite 100,
Pittsburgh, PA 15203

Follow Us

1.877.220.LUNA

1.412.381.5500

getinfo@lunametrics.com

Questions?
We'll get back to you
in ONE business day.