The recent performance of unemployment in Colombia suggests an almost complete reduction
of cyclical unemployment, and the likely prevalence of its frictional and structural
components. This is probably due to a qualitative mismatch between labor supply and demand.
From this last side of the labor market, there has been a clear and persistent lack of
accurate information about its requirements in terms of capabilities and skills. We propose
a big data methodological approach based on help-wanted advertising. Our data base, once
compiled and compared with employment data from the Colombian household survey
proves to be a more accurate and detailed source of information about occupations, skills
and vacancies than any source available to date in the Colombian Statistical System. On
this paper we detail the context and methodological guidelines of this pioneer case study.