Residential location of a serial offender can potentially be predicted by using models created from home to crime site journeys of solved crimes in the area [N. Levine, Journey-to-crime estimation, retrieved 23 October 2003 from http://www.icpsr.umich.edu/NACJD/crimestat/CrimeStatChapter.9.pdf, last visited 1 February 2005]. Aims of this study were: (1) to examine the accuracy of this technique, (2) to explore relation of modus operandi (m.o.) to the distance the crime was committed from home and (3) to analyse whether the accuracy of prediction is enhanced by taking the m.o. into account. Data consisted of 76 commercial robbery series from the Greater Helsinki area. Accuracy of prediction was tested by using leave-one-out technique: the series which the predicting function was applied to was never part of the function used to predict. The functions allowed limiting the area to be searched to 4.7% (Mdn, IQR = 31.0%) of the study area generally, and to 1.0% (Mdn, IQR = 2.6%) when the suspect's spatial behaviour conformed to the circle hypotheses presented by Canter and Larkin [D. Canter, P. Larkin, The environmental range of serial rapists, J. Environ. Psychol. 13 (1993) 63-69]. Significant correlations between m.o. and the length of the journey-to-crime were found, but this information did not enhance accuracy of prediction. Low percentage of marauder style perpetrators in the data gives support to the possible separation of hypotheses of underlying spatial behaviour in instrumental crimes versus crimes of interpersonal violence or arson. Suggestions for development of investigative tools are presented.