Is the Try-On/Return-Rate Correlation Causal?

A dashboard shows: shoppers who used the virtual try-on feature before purchasing have a return rate roughly 30% lower than shoppers who didn't. A product manager wants to cite this number in a board deck as the feature's proven ROI.

1. Explain why this raw correlation is not sufficient evidence that try-on causes the lower return rate, naming the specific confound at play.
2. Propose a concrete method to get a more defensible causal estimate, given that fully randomizing try-on exposure across all shoppers (a true A/B test) may not be feasible for a highly visible, already-launched feature.
3. Suppose the causally-adjusted estimate comes back showing a much smaller effect than the raw 30% — what would you tell the PM, and what would you not conclude from that smaller number?