Good data handling is a prerequisite for good statistical analysis and interpretation. This paper begins by distinguishing types of numerical data. Care for original data and the risks involved in any process of copying data are emphasized. After discussion of the entry of data into a computer file, problems of scrutiny and checking are discussed at length. An attempt is made to suggest what action should follow when such scrutiny discloses anomalies in some values; if explanations can be found, or if clear evidence indicates corrections that are needed, no problem arises, but otherwise various subjective responses need consideration. Advice is offered on the numbers of digits to be retained in computations and reports. A final section presents ideas on the filing of data and analyses from an experiment.