Breaking Up With Your Keyword Data

Clearly, breaking up is hard to do.

Clearly, breaking up is hard to do.

I couldn’t stay silent on this issue anymore. Every time I see an SEO report that includes things like year-over-year or month-over-month data broken down by keyword or non-branded organic traffic, I die a little inside. Why you ask? Because it’s junk data. There’s a concept that tethers statistical analysis called the representative sample. If a sample is representative of the total population you can do analysis on it just as you would the entire population. However, when you carve out all the keyword data from organic visitors to your site who were logged in to Google, used Firefox 14 or Chrome 25 or newer, and who came to your site via the local carousel, what your left with is some mutation of your total data population.

Presentation

Learn more in my presentation about why this is junk data and what you can use organic keyword data for without violating basic scientific principles. I included annotations in this presentation to help you follow along.

Links From Presentation

You can get all the links I shared or referenced in my presentation from this bil.ly bundle.

  • Justin Schmidt

    Great preso today Annie. I’ve been sending my new team members to your site for awhile now as a “read this and you’ll learn a lot of what you need to know” type resource. Keep up the good work!

    • http://www.annielytics.com/ Annie Cushing

      Thanks, Justin! It’s always thrilling to hear my blog is a training resource for marketers. :)

  • http://kidakaka.com/blog Prasad

    Amazing slide deck, I was guilty of fussing over some of these metrics (all the while fooling myself into thinking that I can get the keyword numbers using WMT). Now I just need to tackle my Hippo wisely :-)

    • http://www.annielytics.com/ Annie Cushing

      This is music to my ears! And you’re in good company. I’m sure I made a bunch of work for quite a few marketers. :)

  • Melody Mertens

    Great info!

    Re: #1, Keywords in the context of landing pages – You noted that branded search was driving traffic to a key product page, and that this warrants further investigation. I agree and also find it noteworthy that this may be due to sitelinks rather than the ascribed theory of the product page outranking the homepage. Something to consider! :)

    • http://www.annielytics.com/ Annie Cushing

      Thanks, Melody. It was something I looked into. These queries don’t trigger sitelinks for my client though. And I can see in both GWT and SEMRush that this landing page gets traffic for company queries. It’s because the homepage has no unique indexable text, so their category pages mention the company name more than the homepage does. The client is aware of the issue but doesn’t want to clutter the design with text. And they have text in images on the homepage, wherer they should be using CSS3 in divs for decorative fonts.

      • Melody Mertens

        Interesting. Thanks for sharing! Now if only we could get dev and content teams to make image text visible to the search engines as you describe… it’s tricky stuff!

        • http://www.annielytics.com/ Annie Cushing

          Oh, it’s actually not. I did it on a WordPress site, and I’m not a designer or developer. There are tons of CSS3 tuts that show you how. Send them one.

          • Melody Mertens

            Good idea! I didn’t so much mean that it’s difficult to do as I did it is difficult to get everyone in a large company to change their ways. I’m not sure if we’ve done a demo though…. that would probably help! Thanks Annie :)

  • Alan

    This looks to have a been a great presentation and I’m sorry I missed it.

    You discuss “reporting” of metrics like “organic non brand traffic” is junk science which I completely understand.

    The examples of “what keywords can we use” and associated GA reports seem to me to be more analysis and not actually reporting to a client.

    In this day and age, what do you actually distill down into a client report in terms of performance?

    • http://www.annielytics.com/ Annie Cushing

      Thanks, Alan. I wish I would have had double the time to go through examples of actual metrics I report on for clients. But I highly recommend focusing on organic landing pages, multi-channel funnel reports (that show assisted conversions for organic), and comparing organic traffic to other channels.

  • http://www.skiusainc.com/ SKI USA

    Love the presentation Annie. Informative and Annie humor makes it even more enjoyable. I think the day is not far away when most of the searches will become (not provided) and we marketers won’t even be able to analyze that data. GWT thus is a huge data resource that will gain importance.

    • http://www.annielytics.com/ Annie Cushing

      Thanks! And I think you’re right. Organic keyword data is a thing of the past, but we still have plenty of data to work with that IS reliable and statistically significant.

  • http://www.dougholtonline.com/ Doug Holt

    I agree with everyone else – it was a great presentation. This is the weakest area of our business and something that I’m working on getting my head around. Your presentation helped ease the pain. Thank you!

    • http://www.annielytics.com/ Annie Cushing

      Thanks, Doug! I’m glad I could ease the pain a bit. I think for many I caused more than I eased. But it had to be said.

      • http://www.dougholtonline.com/ Doug Holt

        It did take me a few bottles of wine to get over the initial shock. It’s a good thing I live in wine country :-)

        • http://www.annielytics.com/ Annie Cushing

          Sure, rub it in. :)