Firms use a rich set of incentives including fixed wages, bonuses, threat of firing and promise of promotion. Yet, we do not have a theoretical understanding of how such a mix of incentives can arise. This paper aims to build a theoretical model which describes the incentive mix as the solution to an optimal contracting problem and provides broader testable implications. The basic model has a principal-agent relationship with unobservable effort. The integrative model includes the basic model and three building blocks: job-assignment, learning and human capital. The derived incentive mix is a consequence of the dual role of firing. It is both an incentive and a sorting decive. The model's predictions are tested on firm-level data from a large pharmaceutical company. The broader testable implications beyond the incentive mix are also confirmed by the data.