A Dayton fuse plant paid its foremen for their slowest worker's output, not their fastest

the problem

Untrained recruits froze with fear at the machines and foremen simply fired the frightened and slow

background

what everyone would do

Hire more instructors and run the same on-the-job, sink-or-swim training faster — the method the article points to shipbuilding for already failing at. It doesn't work because the bottleneck was never teaching capacity: a foreman paid on total group output has no reason to spend time on a frightened beginner when firing her and hiring the next name on the list costs him nothing, and a beginner who has never seen a machine tool learns fear from the machine itself before she learns anything a faster instructor could teach her.

what they saw

Carpenter saw that the failure wasn't happening in the training — it was happening in the two seconds before any training began, when a recruit's own belief that she was physically incapable of the work made her unteachable, and in the foreman's pay structure, which rewarded him for having fewer strugglers rather than for making strugglers succeed. Fixing either alone would have left the other in place; fixing both at once removed the two forces actually driving the best recruits out the door.

the move

Carpenter built a training room entirely separate from the shop floor and staffed it only with women instructors, so a new hire's first sight of the strange machinery was another woman already running it — settling, he wrote, the "I could never do that" fear that had been driving off the best recruits before they ever reached competence. He then barred job bosses (line foremen) from firing anyone in their group and tied each boss's bonus pay to the AVERAGE output of the whole group rather than to its total or its best performers, flipping every foreman's self-interest from culling slow learners to coaching them up.

why it works

Seeing another woman already competently operating the machine directly contradicts a new hire's "I could never do that" belief before a single word of instruction is given, which lowers the fear response that was previously causing capable recruits to quit before they learned anything. Tying the job boss's bonus to his group's average output, rather than its total or its best worker's output, removes his incentive to discard slow learners: a struggling new hire now drags his own pay down for as long as she stays weak, so his fastest route to a bigger bonus runs through coaching her up, not firing her. The two mechanisms compound — a calmer, more receptive trainee is also easier for an incentivized foreman to bring up to speed quickly, which is what let production of the same operation nearly septuple within roughly two months.

the payoff

On one 56-point-tolerance aluminum fuse-body operation, 31 women averaging 8 pieces an hour in January 1916 rose to over 25 an hour within four weeks of the training department opening and to over 55 an hour once the pipeline matured — nearly a sevenfold gain. In the roughly 2,000-worker assembly department, two independent engineering firms hired to estimate the plant's ceiling put it at 15,000 complete fuses a day across two shifts; trained women assemblers reached 38,000 a day in a single shift, more than double what the outside experts thought the whole plant could produce.

where it breaks

It only works when the bottleneck really is fear and misaligned coaching incentive, not raw physical or cognitive fit — Carpenter's own account describes reassigning some hires to lighter machines when a job genuinely exceeded their capacity, which the training-room-plus-bonus fix cannot solve by itself. Average-based bonuses also depend on group size staying small enough for one foreman to coach personally — the plant itself capped groups at seven to thirty workers depending on job difficulty; a foreman managing hundreds could not meaningfully raise a laggard's output and would simply be punished for factors outside his control. And it requires supervisors trustworthy enough not to game the average by discouraging a struggler from official records or quietly pushing her transfer elsewhere instead of coaching her, which shifts rather than solves the original culling problem.

what came after

The following year, the U.S. Council of National Defense's Department of Labor bulletin surveying wartime industrial-training methods nationwide carried Carpenter's own contributed account of his plant's training work, placing Recording & Computing Machines alongside dozens of other war plants the federal government judged worth documenting as models for training unskilled labor at speed — independent confirmation that Carpenter's methods were taken seriously outside his own company at the time, not just a claim in his own trade-press article.

references

1. [1]How We Trained 5,000 WomenIndustrial Management (The Engineering Magazine), Vol. LV, 1918archive.org
2. [2]How the Shortage of Skilled Mechanics Is Being Overcome by Training the UnskilledUnited States Council of National Defense, Committee on Labor / Department of Labor, 1918gutenberg.org

keep it

same kind of clever

• 1970 · AldiAldi didn't hire staff to chase abandoned shopping carts across the parking lot — it made getting a quarter back the customer's own problem to solve.
• 1999 · BBK Electronics (步步高, Duan Yongping)Rather than run three growing product lines as divisions that would eventually compete for the same budget and shelf space, Duan Yongping split them into fully separate companies — two of which, OPPO and vivo, are now each other's biggest rivals.
• 1995 · City of Bogotá (Mayor Antanas Mockus)Bogotá answered jaywalking and reckless driving with mimes instead of more tickets