Payments & Transaction Banking · Module 2 of 4
The Routing Call
Four million card payments a month, five authorisation rails, and a decline that isn't an ending: it falls to the next rail and is tried again. Each rail charges its fee on every attempt, win or lose, and approves its own share; and a quarter of the customers who see payment failed close the page rather than wait. So a retry costs a fee and some of your customers, and because later rails are only ever reached by what the earlier ones failed, the running order doesn't decide who goes first, it decides how much work happens at all. Set the order and the depth of the cascade while four meters run live: net contribution, approval rate, average cost per attempt, and a count of the people who walked away. None is labelled good or bad, because this is a decision the real dashboards don't flag. Your CFO asks about the four-cent rail, the cheapest line in the table. Your head of merchant operations wants the approval rate up. Neither executive is wrong: both instruments are calibrated per attempt and the decision is about order. Which table do you run, does the cheap rail belong in it, and what do the two dashboards say about your choice? Then the margin falls, and the question of which rail leads comes back. Order the attempts, not the price list.