As time goes on, our interaction with large datasets tends to be more involved and less intuitive. Most datasets can no longer be visualized in their entirety and we have to interact with them mathematically, making blind data exploration challenging. In this talk I will present some theoretical considerations of data analysis and data complexity, and I will show that performing a "cooling" operation on data can automatically reveal the inherent structure of a dataset and the properties of its underlying population. I will then present a new tool, the Sequencer, which can generically perform such a cooling operation on arbitrary datasets. I will show that it provides a new way to explore complex datasets and restore visual insight.