Incentives Explain Behavior Better Than Character
Why this matters
When something keeps going wrong in a shop, the first explanation is usually a person: the tech who is lazy, the CSR who does not care, the crew that cannot be trusted with inventory. Sometimes that is right. More often, the shop has built a system that quietly pays for the behavior it is complaining about, and a decent person sitting in that seat does exactly what the system rewards. Fire them and the next person drifts the same way within a quarter. Before you decide something is a character problem, check what the pay plan, the dispatch rules and the scorecards actually reward. This card gives the check and works one case where the owner was ready to install cameras over what turned out to be a bonus formula.
The idea and its sources
The management scholar Steven Kerr put it in the title of a 1975 paper that is still assigned in business schools: "On the Folly of Rewarding A, While Hoping for B." Organizations announce they want one thing, then pay for another, and are surprised when they get what they paid for.
The economist Charles Goodhart made a related point about measures: once a number becomes a target that people are rewarded for hitting, it stops being a good measure, because people find ways to move the number that do not move the thing it was meant to track.
And from safety science, the psychologist James Reason popularized a practical test for blame that the investigator Neil Johnston devised. Ask whether another competent person, with the same training, in the same situation, under the same pressures, would likely have done the same thing. If yes, the problem is the situation. It is called the substitution test, and it is the single most useful question in this card.
The tendency it guards against has a name too: explaining other people's behavior by their character while explaining your own by your circumstances. See related: The Fundamental Attribution Error with Customers and Crew.
Three questions before you blame a person
- What does the system pay for? Read the actual pay plan, bonus formula, dispatch rule and scorecard, as written, not as intended. What number goes up when this person gets paid more, praised, or sent home earlier?
- What does the system punish? Which honest behaviors cost the person money, time or standing? A callback charged against the original tech's pay punishes reporting a callback, not causing one.
- What can the system not see? Whatever is not measured is free to sacrifice. If quality is not in the formula, speed will eat it.
Then run the substitution test: if you swapped in your best tech or your most conscientious CSR, would the behavior likely change? If you honestly think it would not, you are looking at a system problem.
Where shops build the wrong incentive
| Rule as written | What it hopes for | What it actually pays for |
|---|---|---|
| Flat hourly pay with overtime | Steady, careful work | Finishing the last job slowly on a Friday |
| Flat-rate pay per book hour | Efficient techs | Rushing, skipped steps, fights over the good calls |
| Next call goes to whoever marks complete first | Fast turnaround | Marking complete before the paperwork is done, or stretching the 3 pm call to avoid the 4 pm drive across town |
| CSR scored on calls handled per hour | Responsive office | Short calls, missed details, no plan offered |
| Callback deducted from original tech's pay | Fewer callbacks | Callbacks reclassified as new calls or handled off the books |
| Warehouse scored on low inventory value | Lean stock | Trucks running out of common parts mid-job |
| Commission on equipment sales only | More replacements sold | Recommending replacement where repair would do (worked in its own case card) |
None of these rules is stupid. Each one fixed a real problem when it was introduced. The trouble is that each one measures one thing and the shop cares about two or three.
The worked case: parts that stopped getting logged
An eight-tech plumbing shop introduced a quarterly bonus for techs: a percentage of each job's gross margin above a target, computed from the job's billed revenue minus labor and the parts the tech logged to the job. The owner wanted techs to price work properly and stop giving away parts.
The symptom. Two quarters later, the parts variance, meaning parts that left the warehouse to restock trucks but never appeared on any job, went from about 3% of parts issued to about 11%. Restocking had not changed: techs requested about the same volume of parts per job as before. Parts were going out and not being billed.
The character explanation. The owner suspected theft, priced cameras for the warehouse, and had two techs in mind whose truck stock looked thin.
Running the three questions.
- What does it pay for? A higher job margin. And every part a tech logs to a job lowers that job's margin, because logged parts are subtracted in the formula.
- What does it punish? Logging parts. Every honest entry reduces the tech's own bonus.
- What can it not see? Parts used but not logged. The formula only knows what is entered.
Substitution test. Would the shop's most trusted tech, under this formula, keep logging every fitting, every tube of thread sealant, every small valve? Plenty would. Some would start skipping the small ones, not as a decision to steal, but because it was tedious, nobody checked, and skipping it raised the bonus. The owner's honest answer was that a reasonable person could drift.
Checking the data against the explanation. If this were theft by a couple of people, the variance would concentrate in their trucks. If it were the formula, it would spread across most techs and concentrate in small, cheap parts that are tedious to log. The reconciliation showed variance rising for six of the eight techs, and most of the unlogged items by count were fittings, sealants and small valves. The two techs the owner had in mind were in the middle of the pack, not the top. The two techs whose variance did not move were the shop's most meticulous paperwork people, who logged everything regardless.
What that rules out: a theft ring. It does not prove nobody took anything, but the pattern matches the incentive far better than it matches a couple of bad actors.
The fix. The shop changed the formula, not the people:
- Job margin for the bonus is now computed using a standard parts allowance per job type, built from the shop's own history, so logging the real parts no longer moves the tech's bonus.
- Parts logging accuracy became a gate. A monthly truck-stock reconciliation must come in within 5% variance, a little above the old 3% baseline to allow for honest miscounts, for the tech to qualify for that quarter's bonus at all.
- The owner told the crew plainly what had happened and why the formula changed, without accusing anyone.
Confirming it. By the end of the next quarter the variance was back to about 4%, close to where it had been before the bonus. Billed parts per job rose, which also meant customers were now being charged for parts that had been going out free.
What would have pointed the other way. If the variance had sat in two trucks while the rest of the crew held near 3%, and those two techs' restock requests had climbed, the system explanation would not fit and a person-level conversation would have been the right next step. The point is not that incentives always explain behavior. It is that they are cheaper to check first, and the check usually answers the question.
When it really is the person
Incentives explain patterns. They explain behavior that spreads across most of the people in the same seat. They are a poor explanation for one person behaving very differently from everyone else under the same rules.
So the reverse test matters too: if six techs on the same pay plan, same dispatch rules and same scorecard behave one way and the seventh behaves very differently, the system is probably not the cause of the seventh. That is when a direct conversation, documentation and the ordinary performance process are the right tools.
Before you change a pay plan
Changing how people are paid touches employment law, and the layers matter. At the federal level, the Fair Labor Standards Act treats most non-discretionary bonuses, ones promised in advance on a formula like the one above, as part of a non-exempt employee's regular rate, which raises the overtime rate for any week the bonus covers (the rules are in the Department of Labor's regulations at 29 CFR Part 778). States add their own layer and some go further: California, for example, requires commission arrangements to be in a written contract and has its own regular-rate rules. Your state may differ. This is not legal advice: before you launch or change a bonus or commission plan, take the written plan to an employment attorney or a payroll provider who handles your state, and ask specifically how it affects overtime and what must be in writing.
How to check your own shop
- List every number someone is paid, praised or scored on. Pay plans, bonuses, dispatch rules, CSR scorecards, the warehouse's targets. Most shops find more than they expected.
- For each, write the behavior that moves the number fastest. Not the intended behavior, the fastest one. If it is one you would not want, assume some of it is happening.
- Look for paired measures. Speed paired with callbacks, sales paired with customer complaints, low inventory paired with stockouts. A single measure gets gamed; a pair is much harder to game.
- Ask the people in the seat. "What does this rule make you do that you think is dumb?" usually gets a straight answer, and it is the fastest audit available.
References
- Kerr, S., "On the Folly of Rewarding A, While Hoping for B," Academy of Management Journal (1975)
- Reason, J., Managing the Risks of Organizational Accidents (1997), on the substitution test
- U.S. Department of Labor, 29 CFR Part 778, overtime compensation and the regular rate under the FLSA
- See related: The Tech Kept Recommending Replacement and It Was the Pay Plan
- See related: The Fundamental Attribution Error with Customers and Crew