Lightning is the most important natural wildfire ignition source worldwide. However, identifying the specific lightning causing a forest fire is a challenging task. The goal of this study is to understand how different methodological approaches affect the association between igniting lightning and natural fires. To this purpose, we combined data on 267 lightning-caused forest fires from Switzerland with data on cloud-to-ground lightning strokes for the period 2000-2018. We searched for the most probable igniting lightning (candidate) among all lightning that occur during the days (up to two weeks) before fire detection in the vicinity (< 10 km) of the ignition point. We tested the suitability of 14 methods that combine different spatio-temporal approaches and selection criteria. Our results show that each method selected different candidate lightning for a subset of the fires, while for other fires identical candidate lightning were selected by different methods. Methods using criteria that combine simultaneously space and time, such as the index A, selected candidate lightning with short distances from the ignition points and short holdover times (i.e., the time between ignition and fire detection). On the contrary, methods that minimize the holdover time selected a great proportion of candidate lightning located at long distances. The majority of the candidate lighting were recorded within 1 km from the fire starting point and in less than 24 h before fire detection. The proportion of positive strokes was significantly higher among candidate lightning than in the rest of lightning, which supports the hypothesis that positive lightning are more likely to ignite a forest fire than negative ones. This study highlights the importance of the methodological choice when searching for the candidate igniting lightning of a wildfire.