Using machine learning to estimate lost demand in a fulfillment chain

A look at how our lost demand estimation has improved over the years.Now, we think the smarter way of estimating lost demand is to build a model that estimates the customer conversion probability as a function of the fulfillment option availability. To estimate the total demand from conversion data, all we would have to do is to sum the conversion probabilities over all customer visits:
As an example, let us say that a customer opened up their app and only saw 3 out of 8 possible fulfillment options available. In our more recent iterations of the lost demand estimation model, we’ve actually opted to build two different conversion models — one that uses fulfillment option availability data from the pre-checkout stage and another that relies on similar data from the checkout stage.

Source: tech.instacart.com