Variable Pay Design: Evidence on Commission, Bonus, and Profit-Sharing

Variable pay (commission, bonus, profit-sharing) is one of the most empirically researched and most poorly applied compensation mechanisms in use. The behavioral economics literature, anchored by Lazear’s “Performance Pay and Productivity” and the broader Gibbons-and-Roberts work on multi-task incentive theory, documents both the conditions under which variable pay produces measurable performance gains and the substantial set of conditions under which it produces gaming, narrow optimization, or no measurable effect. The persistent gap between what the research documents and how variable pay is implemented in knowledge-work settings is a substantial source of compensation program failure.

This article walks through what the empirical literature supports about variable-pay design, where the evidence is strongest, and where adoption substantially exceeds the underlying support.

Data Notice: Effect-size magnitudes for variable-pay interventions vary substantially across studies, industries, and contexts. Findings cited reflect peer-reviewed research at time of writing; specific dollar magnitudes vary widely and should be benchmarked against context-specific data before applying to specific programs.

What the research actually shows

The strongest empirical support for variable pay applies to contexts with three structural features: (1) output is individually measurable, (2) the contract cycle is short enough that the recipient can observe performance feedback within the relevant decision window, and (3) the work is not multi-dimensional in a way that allows the recipient to game the measurable dimension at the cost of the unmeasurable.

Lazear’s canonical study of windshield installers (Safelite) documented productivity gains of approximately 40% when piece-rate pay replaced hourly pay. The study context satisfied all three conditions cleanly: installations are individually measurable, the feedback cycle is daily, and the work has limited gaming surface.

The evidence becomes substantially more mixed as work moves away from those three conditions:

Multi-dimensional knowledge work (engineering, research, design, strategy roles) violates the no-gaming condition. The Holmstrom-Milgrom multi-task agency literature documents that variable pay on a measurable dimension produces under-investment in unmeasurable dimensions, often offsetting the gains on the measurable dimension.
Long-cycle work (annual planning, multi-year product development, research) violates the feedback-cycle condition. Variable pay tied to annual performance metrics often functions as a rationalized base-salary equivalent with weak incentive effect.
Team production dilutes the individual-measurability condition. Variable pay tied to team output produces free-rider dynamics that vary with team size and visibility.
Intrinsic-motivation contexts can show motivation-crowding-out effects, where extrinsic variable pay reduces intrinsic engagement and produces net-negative performance effects. Frey and Jegen’s motivation-crowding work documents these effects in selected contexts.

The implication is that variable-pay program design should explicitly assess whether the structural conditions are met before assuming variable pay will produce measurable effects. The default presumption should not be that variable pay always works.

Sales commission design

Sales is the canonical context where variable-pay structural conditions are met cleanly: bookings are individually measurable, the cycle is short to medium-term, and gaming surface is bounded by deal-discount controls and customer verification. Commission programs are accordingly the strongest empirical case for variable pay in knowledge-work settings.

Commission design within the sales context has nontrivial sub-decisions:

Commission rate (% of revenue or quota attainment). Typical sales commissions run ~5-15% of new bookings for inside sales, ~10-30% of new bookings for enterprise sales, with substantial variation by product margin, deal cycle length, and competitive intensity.
Pay mix (base-vs-variable split). Common conventions: 60/40 base-variable for inside sales, 50/50 for enterprise, 70/30 for technical sales engineers. The pay-mix choice affects the selection-effect on hiring (who applies for and accepts the role) and the retention dynamics under underperformance.
Acceleration. Commission rates that step up above quota (e.g., 1x rate to 100% quota, 1.5x rate from 100% to 150%, 2x rate above 150%) shift the marginal incentive at the binding decision points.
Caps. Commission caps limit upside above defined thresholds. Caps reduce gaming risk on outlier deals but create disincentive effects at the cap boundary.
Clawbacks for cancelled or returned business protect against gaming via low-quality-business bookings.

The empirical literature on sales-commission design is relatively rich and supports differentiated structures by sales-cycle and product-mix characteristics rather than single-template programs.

Bonus design in non-sales contexts

Annual bonus programs in non-sales knowledge work are the most common variable-pay implementation and the empirically weakest. Common bonus structures attach a percentage-of- salary bonus to performance assessment outcomes, with modifiers for company performance.

The empirical effects on individual performance are typically small. The reasons:

Annual cycle is too long for tight feedback loops. Performance assessed in March that affects bonus paid in February has weak salience to daily decisions.
Performance-rating distributions are typically compressed (most employees rated near the modal category), which compresses the bonus differential and reduces marginal incentive.
Multi-task gaming surface in knowledge work produces under-investment in unmeasured dimensions.
Bonus-as-base-salary-equivalent dynamics emerge when bonuses are paid relatively predictably each year to most of the population. Recipients anchor on the expected bonus as part of base compensation, which removes the marginal-incentive effect.

The case for non-sales bonus programs is more often retention and compensation-flexibility (allowing some compensation to flex with business outcomes) than individual performance. Programs that frame the bonus program in retention-and-flexibility terms tend to be more defensible than programs that claim individual-performance incentive effects the empirical literature does not support.

For broader treatment of how variable pay integrates with the rest of the compensation program, see compensation design evidence, and for the economic framing of when more elaborate compensation structures pay back, see hiring cost economics.

Profit-sharing programs distribute a defined portion of company profit to employees, typically as a uniform percentage of base salary (or with modest level differentiation). The empirical evidence on profit-sharing documents:

Modest productivity effects in aggregate, with considerable variance across studies and contexts. The National Center for Employee Ownership research base documents productivity effects in the range of ~2-5% on average, with substantial heterogeneity.
Stronger retention effects than individual-incentive effects. Profit-sharing programs tend to retain employees through company-performance cycles where individual bonuses might disappear.
Alignment effects that are real but harder to quantify. Employees in profit-sharing programs tend to report higher engagement with company-level outcomes, though the productivity translation is variable.

Profit-sharing programs work best in contexts where company-level outcomes are sufficiently variable to produce a meaningful pool, where the workforce composition is stable enough to benefit from the retention effect, and where the alternative-allocation-to-base-salary tradeoff favors variable distribution.

How variable pay fails

Common failure modes:

Variable pay applied where structural conditions aren’t met. Engineering bonuses tied to story-points shipped, research bonuses tied to publication count, customer-success bonuses tied to NPS scores: all produce gaming dynamics that frequently exceed the intended motivation effect.
Compressed performance-rating distributions that reduce variable-pay differentiation to nominal levels. When 70% of the population is rated “meets expectations” with a 10% bonus and 25% rated “exceeds” with a 12% bonus, the marginal incentive is small.
Annual-cycle bonus programs that lack the feedback-cycle responsiveness for incentive effect. These tend to function as deferred base salary.
Bonus-as-retention masquerading as performance. Programs that pay relatively uniformly to most of the population are operating as retention; framing them as performance produces internal-equity friction when performance ratings and bonus magnitudes diverge.

The remediation in each case is honest framing: programs designed for retention should be framed and communicated as retention; programs designed for performance should operate in contexts where the structural conditions support performance effects.

For closely related coverage of how compensation programs fit into the broader hiring system, see career ladder design and skills-based hiring evidence.

Designing variable-pay programs defensibly

A defensible variable-pay program has the following:

Structural-condition assessment before deployment. Does the work satisfy individual-measurability, feedback-cycle, and limited-gaming-surface?
Honest framing of the program objective. Is the primary purpose performance incentive, retention, compensation-flexibility, or company-alignment?
Differentiation calibration that produces meaningful magnitude differences across performance tiers, when performance is the framing.
Gaming-surface monitoring that detects narrow optimization on measurable dimensions at the cost of unmeasurable ones.
Periodic re-evaluation of whether the program is producing the intended effects.

Variable pay is a powerful instrument when applied to contexts where structural conditions support it, and a recurring source of program failure when applied to contexts where they don’t. The discipline is in the contextual fit, not in the program-design template.

Takeaway

Variable pay produces measurable performance effects in narrow structural conditions: individual measurability, short feedback cycles, and bounded gaming surface. Outside those conditions, the empirical effects are weak or negative, and adoption frequently exceeds the underlying support. The defensible program assesses structural conditions before deployment, frames program objectives honestly, calibrates differentiation magnitudes meaningfully, and monitors for gaming dynamics. Sales commission programs are the strongest empirical case; annual bonus programs in knowledge work are typically the weakest; profit-sharing programs sit in between with retention-and-alignment effects stronger than direct performance effects.

Sources

Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology. Psychological Bulletin, 124(2), 262-274.
Sackett, P. R., & Lievens, F. (2008). Personnel selection. Annual Review of Psychology, 59, 419-450.
Lazear, E. P. (2000). Performance pay and productivity. American Economic Review, 90(5), 1346-1361.
Gibbons, R., & Roberts, J. (Eds.). (2013). The Handbook of Organizational Economics. Princeton University Press.
Holmstrom, B., & Milgrom, P. (1991). Multitask principal-agent analyses: Incentive contracts, asset ownership, and job design. Journal of Law, Economics, and Organization, 7, 24-52.
Frey, B. S., & Jegen, R. (2001). Motivation crowding theory. Journal of Economic Surveys, 15(5), 589-611.
National Center for Employee Ownership. (2024). Employee ownership and corporate performance: A research review. NCEO.
WorldatWork. (2024). Bonus programs and practices survey. WorldatWork.