Tumor-derived extracellular vesicles (EVs) have emerged as a promising source of circulating biomarkers for liquid biopsies. However, understanding the heterogeneous physical and biochemical properties of EVs originating from multiple complex biogenesis pathways remains a major challenge. Here, we introduce EV-Ident for preparation of subpopulations of EVs in three different size fractions: large EVs (EV200 nm; 200-1 000 nm), medium EVs (EV100 nm; 100-200 nm), and small EVs (EV20 nm; 20-100 nm). Furthermore, this technology enables the in situ labeling of fluorescence markers for the protein profiling of individual EVs. As a proof-of-concept, we analyzed the presence of human epidermal growth factor receptor 2 (HER2) and prostate-specific membrane antigen (PSMA) in breast cancer and prostate cancer cell-derived EVs, respectively, using three different size fractions at the single-EV level. By reducing the complexity of EV heterogeneity in each size fraction, we found that HER2-positive breast cancer cells showed the greatest expression of HER2 in EV20 nm, whereas PSMA expression was the highest in EV200 nm derived from PSMA-expressing prostate cancer cells. This increase in HER2 expression in EV20 nm and PSMA expression in EV200 nm was further confirmed in plasma-derived nanoparticles (PNPs) obtained from breast and prostate cancer patients, respectively. Our study demonstrates that single-EV analysis using EV-Ident provides a practical way to understand EV heterogeneity and to successfully identify potent subpopulation of EVs for breast and prostate cancer, which has promising translational implications for cancer theranostics. Furthermore, these findings have the potential to address fundamental questions surrounding the biology and clinical applications of EVs.