During the postwar period, many countries have de-tracked their secondary schools, based on the view that early tracking was unfair. What are the efficiency costs, if any, of de-tracking schools? To answer this question, we develop a two skills - two jobs model with a frictional labour market, where new school graduates need to actively search for their best match. We compute optimal tracking length and the output gain/loss associated to the gap between actual and optimal tracking length. Using a sample of 18 countries, we find that: a) actual tracking length is often longer than optimal, which might call for some efficient de-tracking; b) the output loss of having a tracking length longer or shorter than optimal is sizeable, and close to 2 percent of total net output.