Reserving outstanding claims is crucial for insurance industry, in the sense that it influences both the cash balance (via pricing) and the solvency of the company. The main objective of this thesis has been to present and compare four well-established and most applied methods Chain Ladder, multiplicative (Bornhuetter-Ferguson), Chain Ladder/multiplicative combined (Benktander) and Bayesian method (Bühlmann-Straub), as well as the model-based Kaminsky approach that has drawn more interest in the academic world. For each method there will be a description of theoretical basis, followed by analysis of workers compensation with data provided by SpareBank1. In general, the four traditional methods are quite alike. They have similar model assumptions, their estimators can be written in the same form, and in particular they suggest analogous solutions. The Kaminsky method, on the other hand, is built on a more logical setting because it postulates that claim counts are Poisson distributed and delayed with respect to some reasonable probabilities. Assessing estimation risk has been required by the new Solvency II regime. Through comparison of Chain Ladder and Kaminsky reserve uncertainty we will get a better perspective on these two different approaches, in the hope of answering the most challenging question: "Which method is best?". To evaluate the methods qualities, simulated dataset will also be utilized at the end such that future liabilities is not a mystery any more, and we can easily compare the estimates with the true reserve given by the data material.