Using Google Optimize, we split your traffic up randomly amongst two versions of your site A & B, in real-time. Half the visitors will see version A, half of them will see version B. We then measure the performance of A and B separately to determine which version performs better.
Comparisons over time are not tests because they are always typically false. There are too many variables that can change to provide a false number such as different time of the year, different promotion, different inventory, different weather, the list goes on. Month over month comparisons are not performance tests and will not provide with a trusted improvement.
When you run an A/B test all the variables are the same for both version A and B (with the exception of the design) because they run simultaneously on your website. You compare the performance side-by-side of each version at the exact same time – a true test.