Which change in the numbers deserves a search for a cause
Shewhart's memorandum of May 1924 gave an engineer a way to separate ordinary variation from a change worth investigating.
Portraits
The examination is a quota of hours, not an open study
An examiner's time for one application is set in advance by the office, and complexity moves the number of hours rather than removing the limit.
Buying a number with lost revenue
A holdout study switches advertising off in one region for weeks, and the sales given up in that silence are what buy the number.
A number a committee can file away, a shape it cannot
Nightingale and Farr worked from mortality tables that officials had already read and shelved.
The epidemic was measured by subtraction, not by argument
Paris certified 243 influenza deaths in the winter of 1889, while the city's total burials ran far above the average of the three preceding years.
A share of the result means doing work that cannot be shown
Ilarion Kompantsev takes a share of the result instead of a monthly retainer, but cannot publicly show the campaigns that bring him income.
Why Doing Everything Yourself Cannot Replace Quality Control
One missing tag on a client's thank-you page caused Maryna Butovych to stop hiring developers and build websites herself.
How to Tell Name Recognition From an AI Recommendation
Oleksandr Koliesnikov put twelve customer questions to four AI systems before he changed anything, and appeared in none of the 44 commercial answers.
How to Tell Real Standards From Talk About Quality
Olga Medvetska explains, through Vilha's cot, why a rebuilt fixing can be checked while talk about quality cannot.
When the Dashboard Is Green and the Checkout Is Broken
Daniel Bar Shay is building a maintenance practice around the gap between a healthy server signal and a failed customer transaction.
The Shop With Two Beginnings
Serhii Merkulov dates his work from a first attempt to sell, while the public record begins much later with a denim store domain.
The Repair Network That Found Its Limit
Iaroslav Tugai's failed expansion revealed which parts of scooter repair could move online and which still depended on craft.
The Mattress That Stopped Competing by Name
Andrei Spatari kept the manufacturer but changed the label, making service and product identity part of a less comparable retail offer.
Start from the decision, not the library
Olena Khvostyk first tried to pack every numerology method she knew into one product. The build proved unworkable for a solo maker and unusable for a reader who needed one answer. She rebuilt around the decision the person was trying to make.
When the removable denture failed
In 2000 a Ternopil orthopedic dentist delivered a full lower removable prosthesis that moved and injured the mucosa. He rebuilt the case on six implants with a fixed reconstruction and kept that sequence as method.
Agreement before the document
A reproducible six-step sequence for co-founders: define the decision, surface interests, test scenarios, allocate authority, design revision and exit, then write the artifact.
The engineer who asks about the building first
Almost every renovation conversation in Germany opens with the same question: how much will the state pay for? In an office outside Munich, Dmitri Berdnikow runs that sequence backwards, and not out of principle. A measure that fits a subsidy programme perfectly can still be the wrong thing to do to a building.
Forty years to overnight recognition
Katalin Karikó spent decades being demoted, defunded, and dismissed at the University of Pennsylvania for pursuing messenger RNA. What she never stopped doing was publishing it. When the world urgently needed exactly what she had documented, the record was already there to be found.
The statistician Japan listened to
W. Edwards Deming spent thirty years as an obscure statistician in Washington while Japan gave his name to a national prize and an imperial honor. His story is a lesson in how reputation depends on the audience willing to act, not just on the quality of the work.
The doctor who counted deaths
In 1847, a Hungarian obstetrician in Vienna proved with a ledger and a bar of chlorinated lime that his own hands were killing new mothers. The medical establishment rejected him anyway. That failure became a permanent lesson in how expertise has to be delivered, not just possessed.
The Mathematician Who Refuses to Guess
He spent a decade modelling how molten polymers flow through industrial machines. Then he pointed the same equations at Google's advertising auction.
Essays
Those who teach, grow
A wave of research on the "protégé effect" suggests that explaining what you know is not time taken from mastering it. It may be one of the more reliable paths to mastery itself. From a nineteenth-century surgical ward to a psychology lab in St. Louis, the evidence keeps pointing the same way.
The diaspora advantage
Across the OECD, people who have just crossed a border are more likely to start a business than people who never left. The honest explanation is less a story about grit than about closed labour markets, informal credit networks, and migration itself acting as a filter for risk-takers.
The visa that reads your CV
Britain's Global Talent visa admits leaders and future leaders in science, technology and the arts. But the decision that matters happens before any immigration officer opens a passport. An outside panel reads a CV, letters of recommendation and a stack of scanned documents. What counts as proof of recognition turns out to depend heavily on which desk is doing the reading.
From the lab to the market
Most new PhDs used to become professors. Most no longer do, and the numbers on where they land now say less about individual failure than about what doctoral training actually builds. A look at what travels from the lab into industry: the method, the writing, the discipline of peer review. And what quietly stays behind.
What an hour of expertise costs
The same sixty minutes sells for $21 if you are an average freelancer and for nearly $3,000 if you are a senior partner at an elite law firm. A history of the billable hour, and the decades-old rebellion inside accounting and consulting trying to kill it in favor of pricing outcomes instead.
Delegating to machines that talk back
Four field studies (a call center, a writers' desk, a consulting floor, and a software team) converge on the same finding. Raw access to AI barely predicts who gets ahead. What predicts it is whether someone built the habit of checking the machine's work, and learned exactly where to stop trusting it.
Clean data first: fraud detection as a craft
Ad fraud is usually pitched to marketers as a security problem, something to patch and move past. The practitioners who actually keep their numbers trustworthy treat it instead as an ongoing discipline of data hygiene. Every model built downstream inherits whatever contamination gets let through.
Half of advertising is wasted. Which half is now knowable
For a century the advertising trade repeated a line about wasted spending that almost certainly nobody ever said. Field experiments have finally started to answer the question the myth never really posed. They include a search-engine blackout at eBay, a pool of twenty-five randomized trials, and tests by geography instead of by person.
The Vocation Review publishes profiles, interviews and method pieces about people whose professional practice is worth studying. Read our editorial standards and corrections policy.
Updated 3 August 2026.