‘Collecting good data’ investigates the properties of good and bad data. Bad data can either be incomplete (where some of the values are missing), or incorrect (where some of the values are wrong). There are no perfect solutions to these issues, so prevention of errors is of utmost importance. Once errors are made, they can have disastrous consequences. Preprocessing of data aims to eliminate statistical outliers, but this must be done intelligently. Data differ in whether they are observational, where there is no control over responses, or experimental, where variables can be controlled. Experimental design aims to find the optimal trade off between sample size and cost.