At 12:01 this morning, the 10 percent "global tariffs" that have covered most US imports since February expired... and were immediately replaced by new duties of 10 to 12.5 percent on 60 trading partners, covering 99.4 percent of everything the country imports. For most companies, the math barely moves; one non-partisan estimate (the Yale Budget Lab) put the effective tariff rate at around 11.8 percent before this change, and the new duties are expected to add a point or two on top.

What actually changed overnight is the plumbing underneath it all, not the size of the bill. The old tariffs were a stopgap with a built-in expiration date, put in place after the Supreme Court struck down the broader version back in February. These new ones come through a different legal door entirely, a months-long investigation into forced labor , and the administration says it intends them to be permanent. So this is now rewarding behavior, not geography: countries that committed to banning forced-labor imports get the lower 10 percent rate, holdouts get 12.5 percent which is why India's rate dropped this week, after agreeing to adopt the ban. US Trade Representative Jamieson Greer, at a Senate hearing Wednesday: "Our policy remains the same. The specific authorities this administration is using have changed, but the trade strategy has not."

All of this is landing in the middle of earnings season, which means hundreds of companies wrote their tariff talking points last week and are already having to test them against a new set of rules this week. Three different postures are on display. General Motors put a real number on it: CFO Paul Jacobson said tariffs cost the company about $900 million in the second quarter, with the full-year bill expected to land between $2.5 and $3.5 billion and GM raised its outlook anyway. AngioDynamics hedged, telling investors its read on the tariff situation "remains dynamic and subject to change." And Autoliv pegged its numbers to a specific date, noting the guidance assumed tariffs stay as they were on July 9.

What makes this round different from the last year and a half of tariff whiplash: the next moves are already scheduled, out loud, on the calendar. The new 50 percent duties on Canadian autos, dairy, and alcohol take effect August 19. A second investigation into whether countries like China and the EU have built too much manufacturing capacity is due to wrap up in the coming months, and outside economists expect it to push rates even higher. A senior administration official was blunt about the thinking behind all this: the president has two and a half years left in office, and intends to spend them on trade. So you don't have to guess whether the ground will move again because the administration is telling you exactly when.

So before your next leadership call, put a date on your tariff talking point. Something closer to "our numbers reflect tariffs as of today, and there are two specific dates that would change them: Canada on August 19, and a second ruling later this year." That's a sentence your CFO can say with a straight face, and one you can defend to a reporter, and it turns the thing everyone's afraid of (like the ground shifting under you) into something you already told people to expect. The harder call is GM's route: putting a real number on the damage buys you credibility the day you say it, and a public target you'll get measured against the day it's wrong. Whether that trade is worth it probably depends on how much your board and your investors already suspect the number is coming anyway.

BYD, the Chinese car manufacturer, delivered 1,265 cars to Australian customers that were built in 2025 but sold as 2026 models. The company's explanation was that its system recorded the date the cars shipped from the factory instead of the date they were actually built, and BYD insists there was no intent to deceive anyone. Mechanically, the cars are identical with the same hardware, same warranty. The story surfaced in Australia this month, and local coverage credits a frustrated customer's Facebook post with forcing it into the open.

BYD's first fix was a payment of roughly A$1,100 (about US$760). That's when the backlash started, and here's why: a car's model is the number that sets resale value. These are electric cars, which already lose value fast, and a car with a 2025 build date is worth measurably less than an identical 2026 model the moment you try to sell it. On a car worth $47,000 or more, that's worth a lot more than $1,100. Customers felt like they'd paid full price for a 2026 car, been handed one that was already a year old, and then been offered pocket change to make it go away.

The backlash did its work.

Within about a week, BYD expanded into full fix-it-mode: affected buyers can now demand a complete refund of the purchase price, swap their car for a genuine 2026-built replacement, or keep the car and take the cash with BYD covering the costs either way. The company's Australian public relations director apologized publicly and confirmed the expanded offer. Reported potential cost to BYD: up to A$60 million. Same 1,265 customers, same underlying facts, priced twice by the same company within a matter of weeks — once like a paperwork fee, once like a real financial loss.

The difference between the two numbers is the lesson here because BYD's first offer treated this as a clerical mistake worth a small flat fee. But its customers experienced it as a real hit to what their car is worth and the press coverage didn't quiet down until BYD's number matched the customers' math instead of its own.

The advice here is to put this question to the room before any final 'solutio' goes out the door to customers, especially once legal and finance have already agreed on the smallest possible number: does this number treat the problem the way WE see it, or the way our customers will FEEL it? If your number is sized for a paperwork error and the customer's loss is sized like a big financial hit, that becomes the next headline. BYD has been outselling Ford, Hyundai, and Kia in Australia , a newcomer brand running almost entirely on borrowed trust. The full package will probably close out this particular episode but what lingers is the damaged trust.