In order to gain a better understanding of how to improve the quality of small-molecule single-crystal X-ray diffraction data achievable in a finite time, a study was carried out to investigate the effect of varying the multiplicity, acquisition time, detector binning, maximum resolution and completeness. The results suggest that, unless there are strong arguments for a different strategy, a good routine procedure might be to optimize the conditions necessary to get the best data from single scans, and then choose a multiplicity of observations (MoO) to utilize the available time fully. Different strategies may be required if the crystal is highly absorbing, is larger than the incident beam, is enclosed in a capillary tube or is unusual in some other way. The signal-to-noise ratio should be used with care, as collecting data for longer or at higher multiplicity appears to give a systematic underestimate of the intensity uncertainties. Further, the results demonstrate that including poor-quality data in a refinement may degrade the result and, in the general case, the accidental omission of reflections has a very small impact on the refinement as long as they are omitted at random. Systematic omission of reflections needs a convincing procedural justification.