Researchers often evaluate the causal impact of educational practices and programs in observational settings where violations of “no unobserved confounders'' are loosely assumed or disregarded. This paper proposes and integrates two complementary analytic strategies, sensitivity analysis and partial identification, to bound the strength of omitted variables and causal estimates, respectively. The omitted variable bias framework for sensitivity analysis quantifies how strong an omitted confounder would have to be to alter research conclusions. Stability-Controlled Quasi-Experiment is a partial identification strategy to addresses self-selection issues and bounds causal estimates based on assumptions of baseline trends for outcome(s) of interest. This paper demonstrates these two approaches to evaluate the impact of enrollment in a compressed, asynchronous online winter term on community college outcomes.