Diagnosing a Sample-Ratio Mismatch

An experiment configured as a 50/50 split has been running for 6 days. The dashboard shows 61,200 users in control and 58,800 in treatment. Treatment conversion is up 6% relative, p = 0.02. The PM wants to ship.

1. Is the split consistent with a fair 50/50 assignment? Show the check.
2. List at least four plausible causes of the imbalance and how you would investigate each.
3. What do you tell the PM about the 6% lift?