The first step towards managing and optimising the steam network of an industrial cluster or plant is to accurately measure the production and consumption of the different levels of steam. Measurement errors as well as unmeasured contributors lead to inaccurate and open balances. Consequently, operations can be unclear to engineers and costs to managers. Furthermore, steam losses, purges and thermal losses add additional unknowns to an already underdetermined problem. In this study, a flowsheeting tool with data reconciliation techniques is used to model a refinery, closing the steam balances and simulating the steam flows throughout the network. An accurate representation of the refinery’s steam distribution network and its units’ internal steam consumers and producers was the base of the model. Hundreds of online measurements and operator observations gave rise to a redundant and coherent model able to calculate the reconciled steam flows through each node as well as unmeasured flows at each time step. Calculations of steam losses were made possible thanks to a continuous site wide survey of leaks and an algorithm for their accounting. Online and archived measurements were exploited in order to calculate key performance indicators and to verify assumptions over long periods of time. With this method the refinery is able to rigorously allocate energy costs to the steam consumption. A better understanding of the site’s steam network and improvement possibilities is also obtained through reconciled and centralised data and an analysis of key performance indicators as well as steam and thermal losses. The result of this work is also tool for the strategic management of the steam network of the refinery, allowing for the simulation of OPEX improvement scenarios and their related CAPEXs, therefore obtaining a classified list of actions to improve costs.