Data Scouting in CMS

In 2011, the CMS collaboration introduced Data Scouting as a way to produce physics results with events that cannot be stored on disk, due to resource limits in the data acquisition and offline infrastructure. The viability of this technique was demonstrated in 2012, when 18 fb$^{-1}$ of collision data at $\sqrt{s}$ = 8 TeV were collected. The technique is now a standard ingredient of CMS and ATLAS data-taking strategy. In this talk, we present the status of data scouting in CMS and the improvements introduced in 2015 and 2016, which promoted data scouting to a full-fledged, flexible discovery tool for the LHC Run II.