Keyword Segmentation: What Are You Missing In That Pile Of Keywords?

How many unique search keywords drove traffic to your site in the last month? Ten thousand? One hundred thousand? There are so many that they become mind-numbing. We tend to look at the top 25, but below that lies a vast unknown territory. Each of those keywords represents a person searching for something on the Internet, who has chosen your site as the best place to look at that point in time. We don't know the names of those people, and we can only attempt to try to guess at their intentions, but we do know the keyword they chose.

Keyword segmentation is a way to group keywords according to how their owners behave on your site. We do not care if keywords are branded or non-branded; long or short; general or specific. We are just looking at actual performance: did the visitors from that keyword bounce, browse, or buy? Which keywords consistently delivered great customers, versus the ones that had sporadic or nonproductive traffic?

Once we have segmented keywords into groups with similar behavior, we can start to build SEO and paid search plans around those groups. We are doing what marketers have done for decades: segment, understand, and target. Inevitably we will have our super-premium keywords, but what we want to find are the promising niche groups, overlooked workhorse keywords, and the rising stars. Just as politicians are finding voting groups that have been long neglected, and then winning elections, we can find those keywords that can give rise to much better results online if given the right attention.

To do keyword segmentation, we can borrow tried-and-true techniques from traditional marketing. We select all of the performance metrics that characterize keywords, then group them into similar clusters. Ideally we run a statistical model called cluster analysis using a tool like SPSS or SAS. You can try to use Excel to do this manually, but you will need to limit your metrics to just two or three.

A case study

For one retail client, we selected all of the keywords that had delivered at least 10 visitors to the site during a given month. We identified six independent metrics that we considered important to keyword performance:

- Visitors

- % New vs. Returning Visitors

- Pageviews per Visitor

- Actions per Visitor

- Conversion Rate

- Average Order Value

Note that, other than visitors, each of the metrics was an average or ratio, preventing us from focusing too much just on the volume of traffic. We ran the cluster analysis and came up with five core keyword groups, which we gave names to, like "Hotlist" or "Window Shopping," according to the unique characteristics of each:

Metrics Insider chart

As a result of this work, we were able to create an actionable list of keywords that should have a laser-like focus (the Hotlist) as we build out content in the site and buy keywords on the engines. And we could permanently exclude the keywords with no value (Bouncer) from SEO and keyword campaigns. In future months we can nurture keywords in the niche groups depending on the marketing objectives.

So don't ignore the rich information that your keywords are trying to give you. There may be thousands of them, but they can give you clues to what your customers are seeking and where your next opportunities are waiting.

Tags: keywords, metrics
Recommend (258)
1 comment about "Keyword Segmentation: What Are You Missing In That Pile Of Keywords?".
  1. Shilpa Mehta from Mediacom , March 17, 2010 at 11:24 a.m.

    Stan - I completely agree with you on your findings. It is also interesting to look at Keyword User path analysis at unique user level. You can apply confidence interval between the various groups/paths to see the statistical difference. We can know what keywords drive conversions as well as different paths that a user has searched for (converters/non-converters) and that can ultimately help us build a solid group of keywords for search campaigns and SEO as well. Thanks!