There are a lot of electron-beam based techniques in surface analysis, and each of them has its own characteristics, but they also have, at least, one characteristic in common, the information about the target sample is obtained through the analysis of identified signal data. These techniques generally are inefficient for quantitative purpose because only the signal data contribute to the conclusions, while other detected data, the overwhelming majority of measured data, have been completely disregarded as undesirable background data. In this talk, we proposed a data-driven analysis method [B. Da, et al. Nature Commun. 8 (2017) 15629; J. Phy. Chem. Lett. 10 (2019) 5770; Phys. Rev. Appl. 13 (2020) 044055] to extract meaningful information from the background signal and to propose an important breakthrough for the next generation surface analysis. The unique feature of this method is to use the combinations of a large number of spectral groups measured by intentionally changing a plurality of experimental conditions, to describe the background data, instead of interpreting individual spectrum in terms of physically meaningful parameters. Some combinations provided an “intermediate level” between “background signals” and “understandable information,” which enabled a better understanding of measured backgrounds.