A Guide to Developing a Testing Strategy
Web testing is a difficult topic to discuss primarily because very few people know what it is or what it should accomplish. For our purposes, let’s define it as an agnostic approach to implementing changes. Put another way, testing is a method of generating and validating ideas.
This is nothing new to the world of science and engineering. In fact, testing comprises the entire practice of these disciplines. So, to avoid defining a wheel (a sort of circular, smooth object supported on the inside? carries weight somehow; balances . . . let’s save this for a part 2), let’s look at the steps of the “scientific method” – the same one we all learned about in 4th grade:
This is a bone-crushingly simple and effective testing strategy that will work for any site. Let’s work through this as a short exercise, say, for example with a landing page that isn’t converting.
Our testing subject: A landing page that isn’t converting.
- Ask a Question
- Do Background Research
- Construct a Hypothesis
- Test Your Hypothesis by Doing an Experiment
- Analyze Your Data and Draw a Conclusion
- Communicate Your Results
Ask a question:
Why isn’t the landing page converting?
Do background research:
A simple Google search here will bring us to thousands of people yapping about improving conversion rates on landing pages. Here’s where we need to be careful – this is an idea-gathering phase. Try not to be overly focused while doing this – the ideas might be too big or too small. I found some general themes about non-converting landing pages:
- The landing page is too busy, i.e. too many links or too much content
- The landing page doesn’t focus on its conversion
- Your lead capture form is too intense
- Your headlines/value propositions/imagery/etc. are bad
Construct a hypothesis:
Now you would look at your very own stinker of a landing page and generate some ideas about why it might not be working. Depending on your research, you might decide on a complete redesign, or you may try to change only a few aspects, e.g. re-doing the lead capture form or updating the headlines.
Your hypothesis will determine the scope of your test. A complete redesign may require design assets, wireframes, imagery, etc., while changing some elements on the page could be accomplished with a tool like Unbounce.
If you decide that your landing page needs a redesign, your ‘hypothesis’ would read something like this: Our landing page isn’t converting because the design is clunky, outdated, and difficult to scan. Our users probably can’t even tell what the page is about.
This stage of the testing process is where you can get funky – if you’re scrapping your entire landing page, now’s the time to try some radical design variations you may have had in mind. Feel free to ruthlessly copy other landing pages that look good; figure out what everybody else is doing and think about what you like and don’t like. Try some things that might seem silly or too far – it’s just a test.
Even if you’re doing a complete redesign, you’ll need to use an a/b or multivariate site testing tool. Optimizely, Unbounce, Google Analytics, etc. The particular testing tool will be different depending on what you’re testing – most tools are replaceable by others.
Analyze Your Data and Draw a Conclusion:
This is pretty straightforward – which test variations got the best results? eCommerce and lead gen web properties typically only have one or two goals, so there isn’t much in the way interpretation that you might be doing.
Communicate your results:
If that’s what you really want to do . . .
With all of that said,
Testing works as well for solving basic landing page issues as it does for addressing usability issues, information architecture concerns, and validating high level marketing strategies. These are the same steps and concepts that we use to find out whether Coke would dissolve a tooth if left overnight and other essential medical truths. Testing is a process that we use to answer important questions and hopefully find effective strategies. And when you’re through with all of that, you might even have something they call knowledge.