We placed real job advertisements to observe labor market participants’ choices in their natural environment. We used a three stage experimental method. Following a standard job post, applicants had to fill in their credentials and to perform a pre-employment test to measure their ex ante productivity. We then randomized all participants into different treatments varying the flexibility in working hours and the total hours worked, and finally, made real job offers. Our study is innovative in that it has an ex ante measure of productivity that allows distinguishing between the enhanced productivity induced by sorting and the one induced by motivation (or other ex post features). We find that productivity cannot be raised by announcing flexible contracts because more productive individuals do not sort. We do find that flexible contracts attract more women with dependents.