Of course, this article looks at split testing from the point of view of keyword management, but it can be, and has been, applied to all kinds of result-dependent activity.
By result-dependent we mean an activity that has a clearly defined and measurable goal; one which produces results that further our guiding strategy.
In strategic management terms, these are results that keep us in line with our mission and contribute to a sustainable competitive advantage.
At the core is something that I may well have alluded to before: PDCA.
Plan Do Check Act
The PDCA cycle (also known as Deming's Cycle) is something that I first came across in the context of process improvement. For a great overview, try the US Navy's Handbook for Basic Process Improvement, which is part of their Strategic Planning collection of documents.
(It's pretty lengthy, so don't get sidetracked, but there's a lot of useful flowcharts in there that can be easily applied to outsourced projects i.e. manual keyword expansion from a root keyword phrase, and checking of vital keyword effectiveness metrics against baseline measurements, and so on.)
The point of PDCA is that you first make a plan. Then, you carry out a process that has been created to achieve the goal set out in the plan.
Next comes a period of checking -- were the expected results achieved? -- and adjusting the process (the Do) part so that there is a higher chance of achieving the goals set out in the Plan.
In a nutshell, that's the Deming cycle, and it can be applied to a lot of different activities. Such as: A/B split testing of keyword rich headlines that form part of an advert campaign (on AdSense, for example).
However, before we look at A/B split testing (the process, or 'Do' part), we need to make sure that we understand what it is supposed to achieve. For that, I like to use SMART goals.
SMART Goals
There's a great discussion of the various meanings behind the SMART moniker on the Project Smart web site. While you're free to choose a set that make sense to you, here's what I recommend:
- S : specific (i.e. not at all vague)
- M : measurable (i.e. numerically)
- A : achievable (i.e. not unrealistic)
- R : results-oriented (i.e. has to take you to where you want to go)
- T : time-bound (i.e. set a time and/or cost limit)
In this case, your goal might be to increase click-through rates by 1% within 6 months by varying keyword placement in an advert heading.
The goal is set, but what is the process that will deliver the desired result?
A/B Split Testing
The driving principle behind A/B split testing is that you have two test cases, and you want to see which one performs the best. When you find that out, you devise another case and pitch it against the current best performer.
It sounds simple... and it is.
Within our PDCA process, it's the 'Do' part:
- Plan : "increase click-through rates by 1% within 6 months by varying keyword placement in an advert heading";
- Do : execute split testing for up to 6 months (or when CTRs increase by 1%)
- Check : did we achieve the goal?
- Act : if not, how can we improve/change the process?
The point is that Google AdWords will present you with hundreds of keyword phrases, along with data to help you pick the best one. But, it's the best one from Google's point of view, or at a push the best one that's identified using anonymised data.
It might not be the best one for you.
By split testing across candidate keywords, you can easily figure out which gives you the best return on your investment. Here, we've only put the goal in terms of CTR, but it could equally have been oriented around the eventual action of the visitor.
I've also talked about varying keyword phrases, but what about varying the words around those phrases? A great source of variations can be found in a free eBook "87 Marketing Secrets of the Written Word" by the legendary Ted Nicholas (link goes to his main site, and isn't an affiliate link.)
By picking carefully and pitching different combinations of words against each other, you will quickly build up a reliable list of variations for your chosen niche, as long as you follow PDCA, SMART and A/B Split Testing principles.
