Safelite paid per windshield with a floor, output rose 44%

Safelite, the largest US auto-glass company, paid its mobile installers by the hour in the early 1990s. Hourly pay in field service has a known pathology: effort is unobservable (the installer is alone in a van), so pace drifts to the comfortable minimum, and the payroll cannot distinguish a slow worker from one who drew long drives and difficult jobs. Installers averaged roughly 2-3 windshields a day, and management suspected the number said more about incentives than about windshields.

The classical fix — piece rates — was well known and widely distrusted: workers fear rate cuts once they perform, resent income risk from slow days they don't control, and quality can collapse when speed alone pays. CEO Garen Staglin and president John Barlow's design confronted the distrust directly rather than pretending it away.

From 1994 Safelite rolled out 'Performance Pay Plan': installers earned a rate per windshield installed, with a guaranteed hourly floor so a bad draw of jobs couldn't crater a paycheck, and re-installation of any defective work done unpaid — pricing quality into the piece rate. The floor was the adoption device: it removed the downside that makes workers fight piece rates, while the upside pulled effort and — just as importantly — changed who wanted to work at Safelite, as fast installers were now paid like fast installers.

Paid per windshield with an hourly floor and unpaid rework, making the worker residual claimant on pace while the floor defused resistance and selection compounded gains.

Piece rate makes worker residual claimant on pace, so effort is self-auditing.

Guaranteed floor removes income risk, defusing worker resistance to piece rates.

Unpaid rework prices quality into the rate, closing the quality loophole.

Upside attracts fast installers and repels slow ones, compounding via selection.

When effort is unobservable, price the unit and absorb downside risk with a floor; selection compounds the gain.

Output per installer rose ~44% — half incentive effect, half self-selection — pay rose ~10%, profits rose; documented in Lazear's AER 2000 study. Became the foundational natural experiment of personnel economics — the standard evidence that pay design changes both behaviour and workforce composition, taught in every labour economics course.