I don’t really like talking about statistics too much, but I think this topic is really important, I’m talking about clusters. A cluster is my word for a statistical group, a collation or collection of people with similar psychographic, demographic blah blah blah, they’re the same, right? They’re described by gender, age, preferences etc.
Imagine if we could take clusters of keywords instead of people, what descriptors would we use to group them?
As chance has it, we do take clusters of keywords. In AdWords advertising we call these clusters ad groups. So bear with me while we take this into the important bit of this post.
Exactly like taking a cluster of people, we want to break a cluster of keywords down to the smallest group so that each descriptor we use for the group is homogenic (describes the entire group). The best and most complete keyword cluster would be 1 term, phrase or word, but, exactly like a cluster of 1 person, a cluster of 1 keyword is unmanageable because of the volume of ad groups we’d have to make. Because of this we have to group these best clusters into something more manageable.
How do we do this? I’ll use cars as an example (I’m a closet petrolhead, I admit it).
A cluster of car keywords might include “fast car, sports car, best fast car, sporty car” etc. if these keywords were people they would have have similar age, gender, preferences and other descriptors and so it would be safe to put them together as a cluster.
However if we tried to put “family car” into that cluster it wouldn’t fit. It would be like putting someone the same age and gender but with different music tastes into the cluster of people, they kinda look the same and someone not paying much attention wouldn’t notice and it would work… kinda.
The problem we face is that Google pays attention, lots and lots and lots of it, we couldn’t hide that the person is just a little bit different to the others and the same way we couldn’t hide that “family car” is just a little bit different to ”fast car, sports car, best fast car, sporty car” etc. So Google has to broaden the descriptors it uses to explain our cluster, the ad group. It can use age and gender, but not preferences. This broadening of descriptors lowers the niche-ness of our targeting and our quality score – that’s a very bad thing!
What I’m trying to say (a short essay later) is to check your clusters. If they can be broken into smaller, more descriptive bits – DO IT! You’ll spend 15 minutes writing new ad copy, but you’ll thank the time when your CPC goes down.
1 Response to “clusters, statistics and adwords”