ATM decline vs usage-limit reconciliationCards & Interchange

Task prompt

A cardholder has complained through CX in #bugs_in_the_mainframe: they keep getting "limit reached" declines at ATMs even though they're well under their limit and have barely withdrawn anything. Find out whether this is a real, widespread problem, what it's costing us, exactly who and what is affected, and how to fix it. Work the investigation through the five questions below, then give your answer. 1. Scope: How many of these ATM "limit reached" declines are there over the full recorded history, how many of them are actually wrong - the cardholder was genuinely under their applicable ATM limit at the time (in particular where they had withdrawn nothing) - how many distinct cards are affected, and over what date range. Give the wrongful-decline rate. 2. Impact: How much attempted withdrawal volume was blocked by these declines, how many distinct cardholders were hit, and how many customer-days were affected - and how much of the blocked volume is attributable to the wrongful declines specifically. If a decline has no recorded attempted amount, call it out separately rather than treating it as zero. 3. Affected cohort: The exact list of affected cards (with their cardholder/account, the applicable ATM limit, and their real usage at the first wrongful decline), split into cards that had used nothing versus cards that had used some but were still under their limit. 4. Remediation: For each affected card, the concrete corrective action for each wrongful decline - which decline records to reverse or annotate, whether the day's limit needs resetting, and how much the cardholder should now be allowed to withdraw (their true remaining headroom). 5. Root cause and recommendation: What in the limit-counting logic is actually causing these wrongful declines (show it from the pattern in the data), and the specific change that would prevent it - including the tradeoff. End your reply with a single JSON object of the headline figures, using exactly these keys: total_atm_limit_declines (all ATM "limit reached" declines over the full history), erroneous_declines (how many of those were wrongful), distinct_cards_affected (distinct cards across all those declines), distinct_cardholders_affected (distinct cardholders across all those declines), blocked_withdrawal_usd (total attempted withdrawal dollars across all those declines, to cent precision), customer_impact_days (distinct card-and-calendar-day pairs across all those declines).