Your question is Offline Accuracy vs A/B Results. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
If our talent recommendation model has high offline accuracy but fails to improve conversion rates in an A/B test, how would you debug the discrepancy at Andela?