by Arsham Mirshah
What do you remember from middle school science class?
Your teacher would be proud if you could recall that heat expands and cold contracts, except in the case of water where ice actually expands due to the crystal lattice that the hydrogen and oxygen molecules… wait – this is suppose to be about internet marking!
You remember the basics of the scientific method, don’t you? It goes something like this:
- Make an observation
- Form a hypothesis
- Conduct an experiment
- Analyze the results
You’re probably thinking: “Oh yeah! – I remember that!”
You might even recall the last marketing campaign you ran and think: “Wait a second; I actually used the scientific method without thinking about it!”
If that’s the case, I’m honestly not surprised because it’s my belief that the scientific method is just a way of explaining how logic works (bias opinion).
Alright, how does it apply to Internet Marketing?
Good question, let’s use run through the steps by using some examples:
Make an observation
There is no such thing as a perfect website (if you disagree, tweet your site to @arshammm).
There are hundreds of reports & tools you can use in Google Analytics to find areas of opportunity on your website.
To name just a few:
- Goal funnel: “Where in the sales cycle are users falling off?”
- Landing page report: “Which landing pages have the highest bounce rate?”
- Advanced Segments (mobile traffic): “How does mobile traffic engage with the site?”
The idea is to make an observation using a tool like Google Analytics. Once you have the data, you can now…
Form a hypothesis
Creative people, WAKE UP! – This is where you come in!
Let’s assume these observations were made:
- Goal funnel: 67% of people do NOT complete the credit card step.
- Landing page report: The top 5 landing pages also have the highest bounce rates and lower than average time on site.
- Advanced Segments (mobile traffic): Mobile traffic visits navigate to your “contact us” (or “locations”) pages & have a 40% higher bounce rate.
With that data in mind, you might come up with some hypotheses that sound like this:
- Goal funnel: “Since the “secure checkout” badge is in the footer, users might not feel safe. If we move the “secure checkout” badge so it’s next to the credit card input fields, more users will complete the credit card field and complete the sale.”
- Landing page report: “Wow, 2 of those pages don’t have any images! Users might be turned off by all the text. If we break up some of the text with related imagery, they will be more inclined to read the post thus time on site will increase.”– An alternative idea could be: “My prospects need answers in real time, and they only buy after trust is built… let’s try installing a live chat service on our website to see if we can’t increase the number of people we engage with”
- Advanced Segments (mobile traffic): “Since most mobile users navigate to certain pages, they likely want that information to be easily accessible. If we create a simple mobile website that makes navigation to those pages easy and obvious, bounce rate should drop.”
The idea is to form a hypothesis which plans an experiment that you predict will result in better engagement and conversion metrics.
Conduct an experiment
Be grateful for your designers and developers.
In forming your hypothesis, the skeleton of an experiment was also formed. Now it’s time to flesh out the details and ultimately test your prediction.
Here is how you might proceed in setting up an experiment to accurately test your hypotheses:
- Goal funnel: Set up a GWO multivariate test where you test 2 versions of the check out page:
- One is the control page & does NOT have any changes from your current check out page
- The other includes a second instance of the “secure checkout” badge right under the credit card capture fields
- Landing page report: In this case, I wouldn’t set up a GWO split test experiment. Instead, I would simply add images to the posts and make an annotation on my Google Analytics on the date the images were added.In the annotation, write that you added images & include a link to each of the posts, also note that you want to check bounce rate on those pages in a month or so.
- Advanced Segments (mobile traffic): Here you want to create a mobile site that makes navigating the “contact us” or “locations” page super simple. Try Google’s Go Mo(free, easy, as with most Google products (bias)). If you have a contact form you want users to complete, you could use HTML 5 to allow smart phones to know the type of input fields (phone number, email address, etc..) being used, making it that much easier for the user to complete the form. If you get a lot of mobile traffic (more than 1,000 mobile visitors per month), it might be worth setting up an A/B split test to allow half your mobile visitors to see the new mobile site and the other half to see your current website. Or you can send all mobile traffic to your shiny new mobile site (probably what I would do).Make sure to annotate analytics so you know the day you made the switch. This will make the next step really easy.
Remember that valid experiments can be easily duplicated, so don’t get super complicated and do something that no one will understand because chances are you won’t understand it either!
The idea is to conduct an experiment where only one variable changes and all else remain constant. That way you can be certain that your data is clean, at which point you can move onto…
Analyze the results
Numbers don’t lie…
Make sure you give your experiment enough time to gather a statistically significant amount of data.
I understand you’re excited to see the results, I usually am too. No need to bust out your college statistics text books; just gather more data than you think you need.
Only then will you be able to confirm or refute your hypothesis.
That’s the whole point, remember?
The idea is to use data, numbers, facts, and figures to make decisions. There have been several instances where we had to reject our hypothesis and go back to the drawing board. You will be surprised at some of the results you get, trust me!
Ok, you’re done, now do it again!
No, really, do it again.
Take the observations (the data) you gathered from your experiment and make a new hypothesis.
Then test that hypothesis with a controlled experiment.
Now take the observations from that experiment and form a new hypothesis.
Then test that hypothesis… Uh oh, we have entered an infinite loop.
Who knew science was FUN!?
I bet you didn’t think you’d ever hear about the scientific method again, did you?
If you use data & a systematic approach to improving your website, you are guaranteed to see improvements in your conversion and engagement metrics (bias, but proven with numbers)!
Good luck & let @WebMechanix know about your experiment and results!
About the author
Arsham is co-founder of WebMechanix, a
Baltimore-based Internet Marketing company focused on search engine and conversion optimization.
Nothing makes the WebMechanix crew happier than to demonstrate snowballing ROI through objective advice & targeted digital marketing services.