I have observed attempts to utilise cohorts to analyse the impact of a discount promotion on sales, while examining the overall cohort turnover. However, this provides inaccurate data, as the January cohort achieved higher sales than the June cohort. 

Ensure that cohorts are accurately accounted for. Please represent the cohorts in absolute months, ensuring that the cohort is represented as a triangular graph, with equalisation on the right edge. Then, study the values of the metric in question, starting from the required month.