Research library Metrics and incentives
Pay for What You Can Count and You Lose What You Can't.
Holmström and Milgrom showed that incentive pay doesn't just make people work harder. It decides where their attention goes, and the work nobody measures loses.
Bengt Holmström and Paul Milgrom, 1991
What they were looking at
Most jobs aren’t one task. A teacher teaches, and also prepares students for the test. A factory worker turns out units, and also looks after the machines. A salesperson closes deals, and also chooses which deals to chase.
Economics had a good model for paying someone to do one thing. Bengt Holmström and Paul Milgrom extended it to people doing several, and applied it to teacher pay tied to test scores, piece rates versus care of equipment, in-house versus independent sales forces, and how jobs get designed.
What they found
“incentive pay serves not only to allocate risks and to motivate hard work, it also serves to direct the allocation of the agents’ attention” Holmström & Milgrom (1991), p. 25
Pay someone strongly for the task you can measure and their attention moves to it, away from the tasks you can’t. So the right strength for any one incentive depends on everything else in the job:
“the desirability of providing incentives for any one activity decreases with the difficulty of measuring performance in any other activities” p. 26
Sometimes the best answer is no incentive at all:
“an optimal incentive contract can be to pay a fixed wage independent of measured performance” p. 26
A footnote on p. 25 cites a 1989 news story: a ninth-grade teacher in Greenville, South Carolina, caught passing statewide test answers to her students “in order to improve her performance rating.”
Holmström shared the 2016 Nobel prize in economics with Oliver Hart for their work on contract theory.
What it means
Strong pay on a measured task isn’t neutral about the rest of the job. It pulls effort away from the work nobody counts. When important work is hard to measure, the choices are to measure it, to soften the incentive on the easy count, or to split the job so each role has one task you can measure.
Where it shows up in a revenue team
- Strong pay on lead volume pulls effort from lead quality, which nobody measures.
- SDRs paid on meetings booked, account managers paid on renewals regardless of size, marketing bonused on MQLs: each points attention at the count.
- Measure the hidden task. Value per lead, or what each month’s new accounts return a year later, turns the unmeasured job into a measured one.
- Or mute the incentive. A smaller variable share is sometimes the better plan, not the weaker one.