This paper by Derek Beach and Rasmus Brun Pedersen analyses the three variants of process tracing and goes on to develop guidelines for using each of them.
"This paper develops guidelines for each of the three variants of Process-tracing (PT): explaining outcome PT, theory-testing, and theory-building PT. Case selection strategies are not relevant when we are engaging in explaining outcome PT due to the broader conceptualization of outcomes that is a product of the different understandings of case study research (and science itself) underlying this variant of PT. Here we simply select historically important cases because they are for instance the First World War, not a ‘case of’ failed deterrence or crisis decision-making. Within the two theory-centric variants of PT, typical case selection strategies are most applicable. A typical case is one that is a member of the set of X, Y and the relevant scope conditions for the mechanism. We put forward that pathway cases, where scores on other causes are controlled for, are less relevant when we take the study of mechanisms seriously in PT, given that we are focusing our attention on how a mechanism contributes to produce Y, not on the causal effects of an X upon values of Y. We also discuss the role that deviant cases play in theory-building PT, suggesting that PT cannot stand alone, but needs to be complemented with comparative analysis of the deviant case with typical cases." (Beach & Pedersen)
Beach, D., Pedersen, R.B., (ND). Case selection techniques in Process-tracing and the implications of taking the study of causal mechanisms seriously. Retrieved from: http://dpsa.dk/papers/Case%20selection%20in%20PT%20-%20Beach%20and%20Pedersen%20-%202nd%20draft%281%29%281%29.pdf