Modern organizations increasingly rely on team-based structures, where individual performance is interdependent. However, standard compensation schemes struggle to reward cooperation: while team incentives foster mutual help, they induce free-riding, and individual output measures may overlook essential pro-social behaviors. Our study investigates how to integrate peer information—workers’ observations of colleagues’ helping behavior—to better align compensation with actual contributions. We develop a theoretical model featuring a manager and multiple workers, each overseeing their own project while also assisting peers. The model contrasts schemes where the manager observes only overall team output versus detailed project performance and considers incorporating peer information via worker reports (“voice”) or peer evaluations (“vote”). Our analysis shows that while standard schemes fall short of optimal outcomes, integrating peer information enhances output and efficiency by incentivizing cooperation and mitigating free-riding, with absolute measures outperforming relative ones when workers are sufficiently lying-averse. To test these predictions, we will conduct an online experiment on Prolific with approximately 500 participants in March, supplemented by additional treatments assessing cooperation in competitive settings and under social biases.