Facebook s A/B option requires a lot of budget to get a statistically significant test. To be honest, the best is just to take an adset with an audience that's proven, put 3-5 new ad variations in there and let them fight it out and repeat again and again until you find a winner, then take them for scale with other winning audiences.