We present a fractional sensitivity analysis of a competitive microfluidic heterogeneous immunoassay for a small molecule analyte. A simple two-dimensional finite element model is used to determine the fractional sensitivity of the assay signal with respect to analyte concentration, flow rate, initial surface density of binding sites, and antibody concentration. The fractional sensitivity analysis can be used to identify (1) the system parameters for which it is most crucial to control or quantify the variability between assays and (2) operating ranges for these parameters that improve assay sensitivity (within the constraints of the experimental system). Experimental assay results for the drug phenytoin, obtained using surface plasmon resonance imaging, are shown to be consistent with the predictions of the model.