One solution is to gradually move instances to you most likely solution.
But continue a percentage of A/B/n testing as well.
This allows for a balancing of speed vs. certainty
But continue a percentage of A/B/n testing as well.
This allows for a balancing of speed vs. certainty
This is especially useful for something where the value of the choice is front loaded, like headlines.