This chapter concerns the forecasting of crime locations using burglary as an example. An overview of research concerned with when and where burglaries occur is provided, with an initial focus on patterns of risk at the individual household level. Of central importance is evidence that as well as being geographically concentrated (at a range of geographic scales), burglary clusters in space and time more than would be expected if patterns of crime were simply the result of some places being more attractive to offenders than others. One theoretical framework regarding offender spatial decision making is discussed and consideration given to how features of the urban environment which affect the accessibility of places (e.g., road networks or social barriers) might shape patterns of offending. A simple mathematical model informed by the research discussed is then presented and tested as to its accuracy in the prediction of burglary locations. The model is tested against chance expectation and popular methods of crime hot-spotting extant and found to outperform both. Consideration of the importance of different units of analysis is a recurrent theme throughout the chapter, whether this concerns the intended policy purpose of crime forecasts made, the spatial resolution of different types of data analyzed, or the attention given to the dimension of time - a unit of analysis often overlooked in this type of work. The chapter concludes with a discussion of means of developing the approach described, combining it with others, and using it, inter alia, to optimize police patrol routes.