We build and estimate a non-stationary structural job search model that incorporates the main stylized features of a typical job search monitoring scheme in unemployment insurance (UI) and acknowledges that search effort and requirements are measured imperfectly. Based on Belgian data, monitoring is found to affect search behavior only weakly, because (i) assessments were scheduled late and infrequently; (ii) the monitoring technology was not sufficiently precise, (iii) lenient Belgian UI results in caseloads that are less responsive to incentives than elsewhere. Simulations show how changing the aforementioned design features can enhance effectiveness and that precise monitoring is key in this.