Several attempts have been made these past few years to obtain a simplified dynamic model of thermal behaviour of buildings from recorded data. System identification techniques are well known in other fields, such as aeronautics. They have been applied with success to analyse various aspects of building analysis. We applied system identification to the problem of the static and dynamic characterization of building energy performances, in order to write an user friendly software. An important feature of system identification is the choice of the appropriate mathematical model, whose parameters are obtained by a numerical estimation technique. For instance, we can select a lumped parameter model, or an ARMA model, or a state space model. We compared the results and ease of use of these various mathematical models, to select the most appropriate one for building characterization. Artificial sets of data (i.e simulated data) have been produced to focus on problems like correlations between input data, biased air temperature, wind effects, and their weight on the validity of calculated parameters. A second order state space model (2 time constants), based on modal analysis, has been chosen to be set up in a software, which we called LADY. Inputs are sampled recorded inside and outside air temperatures, global solar radiation, and power HVAC. We give a help to initialize the algorithm with "physically" meaning parameters. Main outputs are global U-value of the building, equivalent solar aperture, and time constants, with calculated standard errors. Furthermore, the model issued from the analysis can be used for comfort evaluation or heating consumption calculation under different climatic conditions. We hope a large use of this software to better analyse and understand successes and failures of' system identification techniques in the field of thermal building.