When I arrived, the production team's bonus was built on two things. One was billed revenue. The other was tenure.
Tenure is the one worth starting with, because it's common and it sounds generous. Rewarding loyalty and experience is a decent instinct, and I'd keep a version of it as a secondary component. But when length of service is one of the two main drivers of variable pay, you've built a system that pays people for surviving. Nobody says that out loud. Everybody works it out within about a quarter.
Billed revenue was the better of the two, and it was a black box. Nobody tracked time or task completion, so there was no way to connect a person's month to the revenue their bonus was calculated on. Worse, billed revenue included amounts that had nothing to do with what the person had actually done. So the metric was directionally sensible and operationally meaningless, which is a combination that survives for years because it never looks obviously wrong.
Between them, those two components paid for staying employed and for revenue arriving. Neither paid for work.
Time tracking first, everything else after
The first thing I did wasn't incentive design. It was making time tracking mandatory and enforcing it, with real consequences for non-compliance, over several weeks.
That's an unpopular way to start and it isn't optional. You cannot redesign pay around what a role produces if you don't know what the role does with its hours. Any system you build without that data is a guess wearing a spreadsheet, and worse, the people being measured will know it's a guess long before you do.
This is the first step of the method I use, the TAG method — See the Truth. Implement Actions. Achieve your Goals. Incentive work is where skipping it is most expensive, because a compensation system built on an unverified picture of the work doesn't just fail quietly the way a bad forecast does. It pays out, every month, against the wrong behavior, and the longer it runs the more expensive it becomes to change.
While enforcement was underway, I studied the services themselves: what each one sold for, and how much work it actually took to deliver. At the start that was interviews with staff and managers, collecting everyone's best estimate, because no accurate data existed yet. Best guesses from experienced people are a reasonable place to begin and a terrible place to finish. Once real time data started arriving, I could see which work was genuinely valuable per hour spent and which had been quietly consuming the week.
Work out what a role has to be worth
Here's the piece I'd hand to any owner, and it doesn't require a consultant.
Take a role, not a person. Calculate its grossed-up cost: salary plus payroll taxes and burden, everything it actually costs to employ. Then decide the return that role should generate. That multiple is yours to set and it varies more than people expect, because it depends on what your labor costs relative to what your work sells for. A business running on expensive domestic staff and a business running on offshore contractors will land in very different places, and neither is wrong.
Multiply the two. That gives you what I call the role's Target Contribution: the number the position has to produce to justify existing.
Then compare it to what the person in that role actually does all day. Is the activity generating revenue toward that target? If yes, how much, and does it get there? If no, there are only two legitimate reasons to keep the activity. Either it supports another role's Target Contribution more efficiently than that role could do it alone, or it's a non-revenue function that's genuinely indispensable to client satisfaction.
If neither applies, you've found something. Not necessarily a person to remove — often it's a task that accumulated, a report nobody reads, an approval step that stopped mattering two years ago.
I ran this with the owner in strategy sessions, against activity reports, role by role. It's uncomfortable and it's finite. You can do a twenty-person business in a few sessions.
Then, and only then, redesign the pay
With time data and a work study, I could see which activities actually moved money. So I built a points system: every activity worth doing carried a point value, weighted by what it contributed.
The design principle is narrow. Points attach to things the person controls. Completing work. Completing it successfully. Completing it faster than the standard. Points also come off, for outcomes the person also controls: a valid client complaint, a refund, a chargeback. What doesn't earn points is anything the person can't influence, which is where most incentive schemes go wrong. Paying someone on a number they can only watch produces anxiety, not effort.
The threshold is the part I'd defend hardest. The minimum points required before any bonus paid was set at exactly the role's Target Contribution. Not at what the team had managed last year, and not at whatever the budget needed. At the point where the role covers its own economics.
That does two things at once. Below the line, the business is whole and the person is being paid to do the job they were hired for. Above it, every additional point is genuine surplus, so the business can afford to share it generously without anyone doing arithmetic to check. And because the threshold comes from cost rather than from a plan, it isn't arguable. Nobody negotiates a target derived from what they cost.
Some people didn't make the cut. That's the part of this work nobody enjoys and it's the consequence of doing it honestly.
What came of it
Return on production labor went from 4.7X to 6.6X. I want to be careful about what that does and doesn't prove.
It's revenue divided by production payroll, so two other things moved it in the same period. We raised prices, and the team got smaller — the second for more than one reason, not all of them related to this work. Either would lift that ratio without anyone becoming better at their job.
What isolates productivity is the volume underneath it. Completed engagements went up. Successful completions went up. Complaints and chargebacks went to nearly zero. That happened with fewer people, while the work per engagement increased, because raising the price meant genuinely redesigning what was delivered rather than charging more for the same thing.
More work, done better, by fewer people. The ratio improved for several reasons; that part is the incentive system.
Three rules
Pay for what the role controls. Not for revenue arriving, not for time served, not for outcomes three departments away. If the person can't move it, it isn't an incentive, it's a lottery with a performance review attached.
Set the threshold from the role's own economics. Grossed-up cost times the return you've decided that role should deliver. Budget-derived targets get renegotiated every year because everyone knows they came from a wish. Cost-derived targets don't, because they came from arithmetic.
Fix the definition before the period starts, in writing. What counts, what it's worth, what the floor is, and what the pool is calculated on. An incentive whose terms can move after the work is done isn't an incentive. It's a discretionary bonus with extra steps, and people price that in immediately.
Incentive design fails at the first step almost every time. Not at the math, not at the thresholds, not at the payout curve. At the unglamorous, weeks-long, thoroughly resented business of finding out what people actually do all day, which is the only part nobody wants to fund and the only part that makes everything after it work.