The therapeutic efficacy and in vivo biological function of a glycoprotein is significantly affected by its glycosylation profile. For the development of glycoproteins with therapeutic applications, selection of cell lines producing a glycoprotein with adequate glycoform is crucial. Here, we demonstrate an array-based analysis of secreted glycoproteins for rapid and efficient selection of a single cell producing a glycoprotein with desirable glycosylation. Our approach relies on microengraving and interrogation of glycoproteins produced by individual cells in a microwell array in terms of glycosylation profile as well as the produced amount. On the basis of statistical analysis of the interrogation, single cells which are predicted to produce a desired glycoprotein are selected, retrieved, and expanded. We applied the approach to human recombinant erythropoietin (rhEPO)-producing CHO cells and verified the selection of a single CHO cell that produces rhEPO with a high sialylation degree. Human erythropoietin (hEPO) bearing highly sialylated oligosaccharide was shown to display a much longer plasma half-life, resulting in high therapeutic efficacy. This method may find widespread use in the clonal selection for the production of other glycoproteins with specific glycosylation as well as analysis of the heterogeneity in cell populations in a high-throughput manner.