Every day Public Administrations (PA) provide citizens with plenty of services. Due to different factors, such as the involvement of different human resources or the will to deliver lean and versatile services, the same service can show some variability across different organizations. Log files contain the proof of PA process’ variability thus, being able to analyze logs, can be very helpful both for the PA, in order to establish good practices or contextual rules, as well as for the software house companies that need to analyse and to better customize the software they provide. In this paper, we present a methodology that, using log files as inputs, and based on the so-called TLV-diss$$_{\gamma }$$γ, a parametric dissimilarity measure, allows a data analyst to perform a cluster analysis. This methodology helps both PA and software producers to better understand how services are delivered through informative systems and then to better customize them. We show that our methodology can be used to capture the differences in control flow and components resulting from the log files, and then to better reason on the delivery of public services.