The paper describes a new data-scaling program, AIMLESS, and discusses criteria for deciding the 'resolution' of a measured data set. AIMLESS is used after POINTLESS to determine the point group and space group, and it is followed by CTRUNCATE, which calculates structure amplitudes from intensities. AIMLESS improves upon the earlier CCP4 program SCALA by using a more flexible scaling model and providing detailed statistics for data quality assessment. The program performs scaling to correct for experimental conditions that vary during data collection, such as changes in beam intensity and crystal illumination. It also estimates the standard error of intensity observations and adjusts them to better reflect the true error. The paper discusses the challenges of determining the effective resolution of a data set, noting that adding weak high-resolution data beyond commonly used limits may improve structure determination without harm. It presents tests comparing data-processing statistics with refinement against observed and simulated data, and automated model-building. These tests show that including weak high-resolution data can be beneficial. The paper also analyzes data quality by examining internal consistency and signal-to-noise ratio as a function of resolution. It introduces the correlation coefficient between random half data sets (CC1/2) as a more reliable measure of resolution than traditional R factors. The analysis of anisotropy in data is also discussed, highlighting the complexity of decision-making when data are anisotropic. The paper concludes that the resolution cutoff is a contentious issue, and there is no universally accepted method for determining it. However, tests suggest that extending the resolution beyond traditional limits may improve structure determination. The paper emphasizes the importance of considering data quality and the potential benefits of merging data from multiple crystals to enhance weak anomalous signals. It also notes that the nominal resolution is not a reliable indicator of model quality and that well-measured data at higher resolution can provide more information. The paper advocates for a more flexible approach to resolution cutoffs and highlights the need for further research to address the challenges of anisotropic data and resolution cutoffs.The paper describes a new data-scaling program, AIMLESS, and discusses criteria for deciding the 'resolution' of a measured data set. AIMLESS is used after POINTLESS to determine the point group and space group, and it is followed by CTRUNCATE, which calculates structure amplitudes from intensities. AIMLESS improves upon the earlier CCP4 program SCALA by using a more flexible scaling model and providing detailed statistics for data quality assessment. The program performs scaling to correct for experimental conditions that vary during data collection, such as changes in beam intensity and crystal illumination. It also estimates the standard error of intensity observations and adjusts them to better reflect the true error. The paper discusses the challenges of determining the effective resolution of a data set, noting that adding weak high-resolution data beyond commonly used limits may improve structure determination without harm. It presents tests comparing data-processing statistics with refinement against observed and simulated data, and automated model-building. These tests show that including weak high-resolution data can be beneficial. The paper also analyzes data quality by examining internal consistency and signal-to-noise ratio as a function of resolution. It introduces the correlation coefficient between random half data sets (CC1/2) as a more reliable measure of resolution than traditional R factors. The analysis of anisotropy in data is also discussed, highlighting the complexity of decision-making when data are anisotropic. The paper concludes that the resolution cutoff is a contentious issue, and there is no universally accepted method for determining it. However, tests suggest that extending the resolution beyond traditional limits may improve structure determination. The paper emphasizes the importance of considering data quality and the potential benefits of merging data from multiple crystals to enhance weak anomalous signals. It also notes that the nominal resolution is not a reliable indicator of model quality and that well-measured data at higher resolution can provide more information. The paper advocates for a more flexible approach to resolution cutoffs and highlights the need for further research to address the challenges of anisotropic data and resolution cutoffs.