Off-axis integrated cavity output spectroscopy (OA-ICOS) has generated much interest because it potentially allows highly sensitive field measurements with robust optical alignment. We discuss here design choices involved in design of an OA-ICOS instrument and how these choices impact instrument sensitivity, using as our example the design of the Harvard ICOS isotope instrument, which demonstrates the highest reported sensitivity for mid-IR OA-ICOS (2.4×10^-11 cm^-1Hz^-1/2 at 6.7 μm, obtained during measurements of water vapor isotopologues H₂O, HDO, and H₂  ¹⁸O in the laboratory and onboard NASA’s WB-57 high-altitude research aircraft). We compare the sensitivity of several OA-ICOS instruments with differing design parameters, show how comparisons are hindered by differing definitions of instrument performance metrics, and suggest a common metric of MDA_meas, the fractional absorption equivalent to 1σ uncertainty in an actual measurement, normalized to 1 s integration. We also note that despite an emphasis on sensitivity in the literature, in the Harvard ICOS isotope instrument and likely also similar instruments, systematic errors associated with fitting of the baseline laser power are of equal importance to total measurement uncertainty.