We propose a highly tractable way of analyzing business cycles in an environment with random job search both off- and and on-the-job (OJS). Ex post heterogeneity in productivity across jobs generates a job ladder. Firms Bertrand-compete for employed workers, as in the Sequential Auctions protocol of Postel-Vinay and Robin (2002). We identify three channels through which OJS amplifies and propagates aggregate shocks: (i) a higher estimated elasticity of the matching function, when recognizing that at least half of all hires are from other employers; (ii) the differential returns to hiring employed and unemployed job applicants, whose proportions naturally vary over the business cycle; (iii) within employment, the slow reallocation of workers through OJS across rungs of the job ladder, generating endogenous, slowly evolving opportunities for further poaching, which feed back on job creation incentives. Endogenous job destruction, due to either aggregate or idiosyncratic shocks, is countercyclical and thus raises the cyclical volatility of unemployment, closer to its empirical value; but it also stimulates job creation in recessions, to take advantage of the fresh batch of unemployed, and tilts the Beveridge curve up. OJS corrects this tendency and restores a vacancy-unemployment trade-off more in line with empirical observations.