Only the primary objective informs a modeled variant leader
The primary objective in Google Optimize is the only objective used to identify a winning variant reported via Bayesian modeling – so be careful to identify a primary objective that best measures the impact of your test hypothesis.
Per this Google Optimize help article:
“The experiment objective is used to determine whether or not a leader has been found for an experiment. Additional objectives allow you to measure your experiment against other metrics, however, they do not inform when a leader is found.”
Identify a primary objective that has a fighting chance
Make sure your test hypothesis references a primary objective that gives your test a fighting chance of allowing Google Optimize to identify a winning variant.
For example, if you have an e-commerce site, and you change your product details page to better showcase product information – choose a primary objective that reflects increased page engagement as opposed to a revenue based metric. Certainly include revenue metrics in secondary objectives so you can see related impact for your test.
The thrill of victory
As we noted in our blog post “Creativity through failed experiments“, you can learn from experiments that fail to identify a clear leader. However, when Google Optimize does identify a clear leader, a “sea of green” appearing in the reporting interface is very satisfying.
