A cohort table puts each first-purchase month on a row and the months since then in columns: month 0 is the month of the first order, month 3 is three months later. Each cell holds what that group did then, such as the share that ordered again or the value per customer so far, so reading across a row follows one group of customers and reading down a column compares groups at the same age.
The reason to do it is that averages mix customers of different ages. A store that grew fast last quarter is full of new customers who have not had time to come back, so its overall repeat rate falls even when every cohort is as loyal as before. Comparing like with like, each cohort at the same age, takes that distortion out.
Young cohorts need care. A month a cohort has not lived through yet is unknown rather than zero, so an honest table leaves it empty instead of counting it as nothing or projecting it forward, and an average across cohorts should count only the customers old enough to have reached that month. The customer lifetime value calculator turns the same idea into one figure from your own averages.
In nouz the LTV tab on Insights holds the cohort table, with a Value view in revenue or contribution and a Repeat % view, and Insights, Customers shows each monthly cohort's second-order rate within 30, 90, 180 and 365 days. Both read your whole synced history.