APT Chart, page-1584

Note this analysis assumes zero churn (obviously not feasible). If you assume 30% of the customers acquired in FY17 or prior have left and spread these across the FY19 and FY20 cohorts (so your total active customer numbers are consistent with what the company is reporting), then the numbers line up almost perfectly. You get approximately the same result if FY17 churn is 20% and FY18 is 10%. This is a very long way from 50%.