3 ms·
No mention of A/B testing on randomized regions. When you lack data make some. The problem with policies are people. They don't behave the way the policy expect
by karmakaze 6y ago
No mention of A/B testing on randomized regions. When you lack data make some. The problem with policies are people. They don't behave the way the policy expected. I'm not saying all out extreme A/B but finding the right pparameter values in parallel rather in multiple-week series'.