Understanding the structure-property relationships of solid-electrolyte interphases (SEIs) is critical for the rational design of electrolytes for beyond-Li-ion battery chemistries. However, the nanometre-scale dimensions, chemical heterogeneity and beam sensitivity of SEIs make their structural characterisation exceptionally challenging. Here, we present a data acquisition and analysis workflow for four-dimensional scanning transmission electron microscopy (4D-STEM) applied to beam-sensitive, heterogeneous lithium-metal SEI samples. By combining data dimensionality reduction with unsupervised machine-learning approaches, we perform real-space clustering to enhance signal-to-noise ratios across large datasets. Automated Bragg reflection analysis was then used to identify distinct crystallographic phases within the SEI. This approach enables nanoscale phase mapping of dominant species across large fields of view with nanometre-level lateral resolution. We identified lithium oxide and lithium hydroxide as the predominant crystalline phases in the lithium-metal SEI.