Measuring a New 3% US Tariff on Mexican Imports and Re-Forecasting Payments Revenue

A fintech company provides cross-border payment services to businesses. Assume it earns revenue from fees and foreign-exchange (FX) margins on the money its business customers send and receive across borders, and that some of its customers trade between the United States and Mexico. In this scenario, five weeks ago the United States imposed a 3% tariff on goods imported from Mexico. Work through the questions below as a data scientist advising the business. ### Clarifying Questions - Which of the company's flows touch US–Mexico trade: US importers paying Mexican suppliers, Mexican exporters collecting from US buyers, or both? - What customer-level data is available: payment corridors, currencies, counterparties, industry codes? - Was the tariff announced before it took effect, so that importers could have bought ahead? - How is revenue forecast today: by customer segment, by corridor, or as one total series? - For Part 4, is the 3% and 7% an absolute percentage error on monthly revenue or a signed bias, and over what horizon? ### Part 1 — Measure the impact on the business How would you measure the effect of this tariff on the company's business? ```hint Who is actually exposed Use payment records to find the customers whose flows cross the US–Mexico border, and fix that list using behavior from before the tariff. ``` ```hint What would have happened anyway Five weeks of post-tariff numbers mean little alone. Ask which customers or series show what exposed customers would have done without the tariff. ``` #### What This Part Should Cover - A precise, pre-tariff definition of exposure - A credible counterfactual and comparison group - The metrics to measure and the confounders to control (seasonality, currency moves, buying ahead of the tariff) ### Part 2 — Direct and indirect effects The interviewer then asked about direct versus indirect impact: for example, importers may switch to suppliers in other countries. How do you account for substitution and cannibalization? The evaluation must cover the company's whole business, not only the customers the tariff touches directly. ```hint Where does the volume go If a customer moves a supplier from Mexico to another country, one corridor falls and another rises. Ask which unit of analysis sees both. ``` #### What This Part Should Cover - The direct and indirect channels of impact - Net versus gross effect, including volume that moves between corridors - Revenue effects through different take rates by corridor, and effects on customers not directly exposed ### Part 3 — Re-forecast revenue under a lasting tariff Now assume the tariff will stay in place for a long time and will not be withdrawn soon. How would you redo the revenue forecast? ```hint What the old model has never seen A model fit on history has no tariff in it. Decide which forecast drivers the tariff changes, and how much five weeks can tell you about each. ``` #### What This Part Should Cover - A driver-based forecast structure into which the tariff enters explicitly - How the size and timing of the adjustment are estimated - Scenarios, uncertainty ranges and an update cadence ### Part 4 — Investigate a larger forecast error The new forecast is in use, but the error between actual and forecast revenue is now 7%, where it has typically been about 3%. How would you investigate why the error grew? ```hint Break it down the way it was built Split the error along the forecast's own segments and drivers, and ask whether segments the tariff never touched also got worse. ``` #### What This Part Should Cover - The error definition: bias versus noise, and whether 7% is outside the normal range - Decomposition of the error by segment and by driver - Separating wrong assumptions from model error and data problems, and the fix ### Part 5 — Brief the CEO Finally, present to the CEO: explain how you evaluated the tariff's impact and how you would explain the forecast's performance. ```hint Answer first The CEO needs the size of the impact, how sure you are, and what to do about it before any method. ``` #### What This Part Should Cover - A bottom line with a range and a confidence statement - The method and the forecast miss explained in plain language - Recommended actions and what will be monitored ### What a Strong Answer Covers - Exposure defined from transaction data rather than industry guesses alone - A credible counterfactual with explicit threats: buying ahead, seasonality, currency moves and a short post-period - A net, company-wide effect with substitution and cannibalization separated from the gross corridor decline - A forecast that models the tariff through its drivers, with scenarios and intervals - A disciplined error diagnosis and an executive narrative that leads with impact, confidence and actions ### Follow-up Questions - Suppose importers rushed orders through before the tariff took effect. How does that change your five-week read, and how would you correct for it? - If the Mexican peso moved sharply against the dollar in the same weeks, how would you separate the currency effect from the tariff effect? - Which leading indicators would show early that customers are moving their sourcing to other countries? - The CEO asks for a single number. What do you give, and how do you convey the uncertainty around it?