SEO: Long Tail Testing
Early on in SEO I learned traffic to individual keywords is bound to be disappointing. I remember my first client, a company that makes folding bikes named Breezer (yes and I just gave them anchor text and no they are not paying me. Hey, they gave me my first gig, I owe them something!). We got their keywords to the first page and the traffic was yawn worthy (despite what WordTracker predicted).
And then over the years I have ranked for some of the best individual keywords in search (“Airline Tickets”, “Used Cars”) and while the traffic was very nice, it was not game-changing. Besides, there are a precious few of those keywords and it is harder than ever to target specific keywords. Game-changing has always been the Long Tail. That’s what I learned by studying TripAdvisor’s tactics many years ago.
Most of my SEO career has been in-house working with large organic sites with many millions of pages. I have usually been in situations where a huge percentage of traffic comes in from a few templates or page types. I know many SEO’s who attitude towards a page that is working is “Don’t touch my page, ever! I need those links, those anchor text, that boilerplate, don’t screw with my h1’s … yada, yada, yada”.
Ah, but all things must pass and all pages must change! Any good sites must move forward with designs that no longer look like 2008 or 1999! Besides, I want to change things myself, maybe some tinkering can get me another 10% or more of traffic. The problem is, with so much traffic tied to a single page type, change brings risk. And it still takes a long time for Google to refresh a really deep section on a site so if you make a change which hurts the site, you need to live with it for a while.
One of the things I have learned from my SEM brethren over the years – Always Be Testing (ABT). However, I can’t put 30% of my traffic at risk. So when a designer or product manager wants me to reduce text repetition on a page rather than say “Keep you hands off my Text”, I say let’s test. What I do then is fairly simple.
- I set it up so that a subset of my long-tail pages get a test version, perhaps 10% or 20% of the pages. I have the engineering team build some infrastructure so they can implement the rules on which pages get which version (perhaps alphabetic, or geographic or even random). I also have the ability to run multiple versions.
- I trigger an event in Google Analytics associated with each version. I use the event to count landings and also set-up segments to analyze traffic and conversions.
- I also set keyword group for each test and control and monitor rankings over time.
There are many reasons to test and besides satisfying external wishes for design and product changes, I test to boost SEO: Title changes, h1 changes, attempts to isolate specific ranking factors.
Of course some things are harder to test (specifically the impact of linking structure and site navigation), but this is great for on-page ranking factors. You have to give your pages enough time to get re-cached by Google so don’t forget to check Google’s cache before you start analyzing results.
So, if you want to tinker with your tail – TEST! In future posts I’ll give specifics on specific things I have tested.