Abstract. This paper applies a recently developed technique for deriving long-term trends in ozone from sparsely sampled data sets to multiple occultation instruments simultaneously without the need for homogenization. The technique can compensate for the nonuniform temporal, spatial, and diurnal sampling of the different instruments and can also be used to account for biases and drifts between instruments. These problems have been noted in recent international assessments as being a primary source of uncertainty that clouds the significance of derived trends. Results show potential “recovery” trends of ∼ 2–3 % decade−1 in the upper stratosphere at midlatitudes, which are similar to other studies, and also how sampling biases present in these data sets can create differences in derived recovery trends of up to ∼ 1 % decade−1 if not properly accounted for. Limitations inherent to all techniques (e.g., relative instrument drifts) and their impacts (e.g., trend differences up to ∼ 2 % decade−1) are also described and a potential path forward towards resolution is presented.