How to estimate click through rate lift for a ranking model?

Suppose I want to build a ranking model for product search in an eCommerce site. I can use a pointwise or pairwise approach to build a ranking model based on the click data collected under the production model. Then use MRR or NDCG to know if the new model is better than the production model. But how to estimate the click through rate(CTR) lift of the new model over the production model? The pointwise approach produces the prediction for CTR, but I don’t know how to use it to get the lift estimate. The pairwise approach just produces ranking scores, so obviously it is even harder to get the lift estimate for a pairwise ranking model.