Using Google Optimize to play with brains (err, experiment on your site)
I don't know if y'all have noticed, but I've been secretly PLAYING WITH YOUR BRAINS!
Joking, I've been running A/B tests on site using a tool by Google called Optimize and it absolutely rocks. I think all of you should try it.
I ain't even going to try make this a formal tutorial, I'll get right into it without any fluff.
Why tho?
Alright Serg, I got it, it exists. Why should I use it? I've got a tiny MVP and doing A/B testing is pretty stupid, no?
WRONG!
The more data you can gather on how your users behave, the better decisions you can take about product (what to ship and what to stop shipping). Having empirical measurements on how your feature ships/copy versions are faring is CRUCIAL to the learning loop encouraged by the lean startup methodology.
You can test things like…
- How marketing copy affects goal conversion rates
- How your feature ships are increasing time-on-site (that's what I tested this week!)
- How different wording in a button can change how users behave
Okay, so how do I do it?
First of all, prepare to measure right. It's not just throwing the script into GTM and calling it a day.
- Set up goals on Google Analytics. Make sure to track things that users do on-site. Set up goal completions, track events around the site (on ML for example task-create, discussion-start, reply-create, etc). This allows you to gather more specific data tailored to your product.
- Now… throw that script in there! (+ use the anti flash script… or not) Optimize comes with a script that essentially blocks page render until your experiment has been rendered… It's aight but adds substantial load time implications. I don't use it tbh.
- START TESTING! Create some experiments. Use the editor to change some copy on site. See how users react, set up targets to goal completions… You can see how completions increase and after a couple of weeks the results will be emailed to you. Make sure to give the experiments time: it helps gather more data.
You can do various experiment types: A/B tests of different variants on site, redirect tests that take x amounts of people to another site (for example a staging site). You can also select how many people get these experiments (50/50 split, or use GA data to decide based on any criteria you wish).
Aight what's the link?
This was a quick tutorial. Enjoy and ask any questions below. SHIP FOLKS! 💪