DuPont beat Trian's slate in May 2015, lost its chief executive in October, and announced the breakup Trian had sought in December. Winning the vote decides who executes the demand, not whether it happens.
The loudest anger in the set, Nike's, cost no measured sales. The quietest partner revolt, Carrefour's, emptied five countries of shelves. Loud and lost are different numbers, and only one of them is read in the first week.
Uber paid its intruders $100,000 and told no one for a year; the settlement for the hiding was $148 million and its security chief was convicted. Capital One told 106 million people in ten days. Telling has a price. Hiding has a fee, and a bet behind it.
Red Bull holds about 13 percent of the energy-drink market it taught to exist, and Beyond Meat's category belongs to someone else. The market you make is not the market you keep, and the difference is the number no plan writes down.
Live Nation settled with the Justice Department in March 2026 for eight more years of its consent decree and no money. An undertaking is priced by what it changes in how the company runs and for how long, not by the fine.
Sanofi paid $74 a share plus a contingent right for Genzyme, watched the right expire at zero, and then paid more than $315 million to settle the fight over it. A missed earn-out is not a free one.
Square's contracted share of interchange on Starbucks transactions was below its cost to process them, and by October 2015 the losses exceeded $71m. What the logo brought in was never counted, so the concession could not be judged, only regretted.
Walmart lost about $1 billion pre-tax leaving Germany in 2006 and sold its Korean stores for about $882 million the same year. Leaving a country has a price with a sign, and the loss a country reports is not the loss that leaves with it.
Stellantis promised the market €5 billion a year of synergies and reported €8.4 billion for 2023, two years early. In 2024 its revenue fell 17 percent and its chief executive resigned. A promise met and a deal that paid are different findings.
Micron's chief executive died on a Friday. By Saturday the bylaws had made his president permanent, days after that man had announced his own retirement. A plan fills the seat cheaply. The relationships are the half it never touches.
Amgen sued on two patents claiming every antibody that works the way its drug does, and a unanimous Supreme Court struck them in 2023. A strong case is not a decision to sue; the fight is priced against the calendar and what else the patent holds off.
Coca-Cola took back the whole Dasani range the day after bromate was found above the limit. Merck kept selling Vioxx with a label for two and a half years and paid more than $6.6 billion. A recall is priced by how much comes back, and by the day.
P&G organized ten categories into six units in July 2019, each with its own P&L, and folded the sixth back into a fifth twenty-six months later. What those twenty-six months cost is in no filing.
TSMC's first Arizona fab was announced at $12bn and reached $20bn, a year late, and it held. Stanley Black & Decker's Fort Worth plant never ramped and closed with $44m of tooling written off. A reshoring case is argued a unit at a time and decided a plant at a time.
Tesco said its half-year profit was overstated by £250m twenty-four days after the update that carried it, and paid about £235m all in. Steinhoff said 'irregularities' with no figure, and the market priced the silence at more than $12bn in days. The bill tracks who says the number first, not how big it was.
Facebook rebuilt its app in native code in a year and called the old one its biggest mistake. Boeing re-engined a 1960s airframe for being cheaper and faster. The estimate is not what decides a rewrite; what carrying costs in full and the company's own record do.
Ericsson made one Philips plant the sole source of its radio-frequency chips; a minor fire in March 2000 cost it an estimated seven million phones. Nokia bought the same chips from the same plant, redesigned them for other fabs, and took the market. The difference was a count of days.
Southland paid $77 a share against a $70 valuation to take itself private, and defaulted on $1.8 billion three years later. Dell left at $13.88, borrowed $45.9 billion more out of the light, and came back in five years. The debt was not the difference; what the debt left was.
Unity announced a per-install fee on September 12, 2023, was revising it ten days later, and canceled it a year to the day. A change in the unit of billing is paid for by the customers you already have, in the year it lands.
Schwab cut its online commission to zero in October 2019, at 3 to 4 percent of its revenue against 15 to 16 percent of TD Ameritrade's, and agreed to buy TD Ameritrade 54 days later. Both sold the same trade at the same price. What differed was what each customer was worth beyond it.
Starbucks bought every store in China for $1.4 billion and its share fell from 34 percent to 14. Apple owned every store and built the most productive retail in the country. The structure did not decide either; what the outlet sells in each pair of hands did.
Southland's board was right that $77 a share beat the market in 1987, borrowed about $4 billion to pay it, and handed 70 percent of 7-Eleven to its licensee three years later. A low bid and a bid worth refusing at any cost are different findings.
WeWork filed at a $47bn private mark in August 2019 and public investors valued it as low as $10bn; Instacart's 2021 peak met a $9.9bn public price in 2023. The market prices the filing, not the mark, and a listing is priced against what a private buyer would pay instead.
When Apple licensed the Mac to clone makers, around 99 percent of clone sales went to its own customers, and Apple ended it in 1997. Google gave Android away and won. The rate never settled it; whose customers the licensee sold to did.
Walmart handed over half of its Mexican venture in 1992 and paid about $1.2bn for a majority five years later. It handed over nothing in Germany and left at a loss of about $1bn. The share is priced twice, and the second price is the larger.
Meta said Reels was negative to revenue by about $500 million in a quarter and kept going; Blockbuster ended late fees, one dollar in six, and its board removed the man funding it. The same answer, and the difference was never the product.
Shopify paid $2.1bn for a fulfillment network in May 2022 and sold its whole logistics business for a 13 percent stake thirteen months later. Amazon built the same work back over a decade and held. The difference was not whether the work was cheaper inside.
Quibi raised $1.75 billion before it had a subscriber and stopped streaming 7.8 months after launch. Airbnb took $2 billion on distressed terms, signed before it was announced, and listed eight months later. The months a raise buys are worth what they reach.
Netflix told a hundred million households to pay and added thirteen million members in a quarter. Costco raised its fee five dollars and lost seven tenths of a point. The discount you are about to approve is paid to the customers who were never leaving.
Intuit pointed Mint's 3.6 million users to Credit Karma the day it announced the shutdown, and the ending held. Nokia announced Symbian's ending before a replacement shipped and lost more than half its smartphone share in a year. The difference was not whether the product paid.
The New York Times' bundle subscribers grew 44.2 percent in 2024 while the revenue from each of them fell 6.7 percent. A bundle is a price cut that buys years, and the years are the half no filing carries.
Yum announced a recapitalization to about 5.0x EBITDA in October 2016 and carried it as stated. R. H. Macy borrowed against itself in 1986 and filed for Chapter 11 in 1992. Same ratio, different earnings under it, and a recession that tested one of them first.
Spotify has 675 million monthly users and 263 million of them pay. The famous forty percent is a ratio of two stocks, not a conversion rate, and no filing says what share of the free started paying in any year.
AMD bought world-class silicon in October 2006 and Nvidia shipped CUDA a month later; eighteen years on, one has 5.9 million developers and the other paid to run its rival's code. A platform is the layer, not the asset under it.
Schwab's 2019 commission cut cost Schwab about 4 percent of revenue and TD Ameritrade about 16, and the war lasted 54 days. A price war is priced in each side's own revenue, or it is not priced at all.
Microsoft gave fifteen years of cloud streaming to Ubisoft and closed a $75.4bn deal on its terms, because cloud streaming was not why it was buying Activision. A demand is priced against the reason for the permission, not against the deal.
Trader Joe's stocks roughly 4,000 items and turns an estimated $1,750 of merchandise per square foot, more than double Whole Foods. It does not describe that as cost control.
WeWork's prospectus set $47 billion of lease obligations against $4 billion of committed revenue. Both numbers were disclosed. Only one of them was the headline.
Notice is not comfort. It is time to start building, and it is only worth anything to a company that would actually start.
Ford held 72 percent of the electric-pickup segment in the fourth quarter of 2023 and halved the plan anyway. Winning the segment does not fill a plant built for a bigger one.
A program can be up twenty percent on what it paid per share and still have cost the company billions. That is the difference between the comparison that was made and the comparison that mattered.
Richemont's own report gives 10 percent of the equity and 51 percent of the voting rights at 31 March 2024. Both figures are public. They are eleven pages apart.
The saving was real in every case in the evidence, Circuit City's included. What separates the outcomes is whether anybody priced the reversal before it was needed.
Nothing about waiting makes leaving cheaper. That is arithmetic rather than opinion, and most rooms have never seen it stated.
Boeing has taken well over $7bn of cumulative charges on the KC-46, a fixed-price contract signed in 2011, and has kept delivering throughout. Keeping going is not a failure to decide.
The question was never whether both rose. It is whether they rose together — and Pinterest published 553 million users and $2.12 of quarterly revenue per user on the same day.
Amazon's parcel volume passed UPS and FedEx by 2023. It was still a very large customer of both. Nobody sent a notice when the years ran out.
The advantage transferring and the attempt working are different events. Google's distribution transferred perfectly to 540 million accounts, and ninety percent of sessions lasted under five seconds.
Southwest declined for years to replace a crew-scheduling system its own union had been warning about. The December 2022 meltdown canceled roughly 16,700 flights at about $1.1bn. The replacement had been costed.
Buy a dollar of profit at eleven times and carry it at seventeen, and six times that dollar appears on the day — before integration, before a single synergy.
The argument about whether the thesis is right cannot be settled today; the argument about how much you spend before anybody outside acts can be settled this afternoon.
Whether the substitution happens is a question about your customers. Whether you should supply it is arithmetic. Merging them is what makes the decision feel like a test of courage.
A crisis is priced on the gap between knowing and acting, not on the defect. The gap is measured from a document you already hold, and one more quarter of it has a number.
A cross-subsidy is a capital allocation nobody made, and the two numbers that would let somebody make it are already inside your own reporting.
Going direct does not remove the functions the channel was performing. It moves them onto your profit and loss, usually at worse efficiency.
A proposal containing a saving and no cost has assumed a decay rate. It has just never said so, and nobody has been asked to own it.
Lock-in is not what a customer would pay to leave. It is what they will bear before they start looking — and only the gap between those two numbers is collectable.
A portfolio of capable pilots and an empty production estate is the normal position in enterprise AI, and it is rarely a modeling problem. Find the gate that is actually open, and price what a completed unit costs once somebody has to check it.
A flywheel makes acquisition cheaper by a fixed proportion, quickly, and then it stops — which is a valuable thing to own and a poor thing to forecast on.
Control is a purchase renewed every year, not a possession dilution takes away. The price is computable, and the structures that fail almost never fail on it.
A reckoning is a financing problem with a legal trigger. The objective is to turn an open-ended exposure into a schedule you can pay, and what that conversion is worth is computable.
A loss leader is a loan, made at the point of sale, to every customer who walks out with one — including the ones who will never buy anything else.
Expansion fails at the rate, not the market. The rate limit is two numbers divided, and a plan can be comfortably funded and impossible to run.
A moat is not something your business has. It is the gap between what a challenger must spend to take your customer and what that customer earns them, and it is measured in their numbers, not yours.
The lure, the engine and the meter. The first two are yours; the rate the earning scales on usually is not, and that is where money machines end.
A price change moves a number. A pricing model change moves four layers and creates a trough somebody has to fund. Size it before you announce the change, not after the first quarter that misses.
A price rise is decided by two numbers, and almost every company brings only one of them to the meeting.
A doctrine is cheap in every year anybody has looked at it and expensive in the one nobody has, which is the year it gets abandoned by people who never decided to abandon it.
An explanation with no ceiling explains whatever it is asked to explain, which means it can never be wrong — and one that can never be wrong has been agreed rather than found.
One cost is paid once and is visible. The other is paid every year and is invisible. The room compares one against nothing, and holds.
A story with no terms in it is not a decision anybody made. It is an anecdote wearing the clothes of one, and it cannot be learned from because there is nothing in it to check.
The sum of the parts is the only figure in a separation paper that nobody can check. Everything else is in the accounts, the market, or the ledger.
The toll is not collected after the war. It is part of what an adopter is comparing, before anybody has adopted anything.
A turnaround that gives up nothing is a cost program wearing a turnaround's name — and the plan can be right and still be replaced before anybody finds out.
Nobody decides to become impatient. The window shortens by drift, in a conversation nobody scheduled, among people who would each say they had been patient.
You can outsource anything whose interface you can specify and nothing you cannot — and what you give away unspecified, you buy back at a price somebody else sets.
Every adjacency is sold as leverage. This pack separates the strength you can genuinely lend a new business from the strength you only believe you can, and prices the difference.
Asset-light is argued as a multiple story and decided as an operating one. This pack prices the fee stream against the owned economics, and asks where the risk actually goes.
A shrinking core is five different problems wearing the same symptom. Diagnose which one you face, choose the response it implies, and price how much time you can actually buy.
A pack is a kit for running a decision, not a research report and not a course. It is built for the case where a decision is in front of you and you would otherwise start from a blank page. Preparing the equivalent takes twenty to twenty-five hours, or thousands of dollars to commission.
The price does not carry customization, implementation help or support. Worked examples use a fictional business, labeled on every file that contains one, with numbers tuned to teach rather than to flatter.
Where a pack reasons beyond its evidence, the page says so. Each one carries a provenance register listing what is sourced, what is derived, what is Stratrix analysis, and what is not asserted.