The paper looks at the importance of the true banks business model in shaping the risk profile of financial institutions, and therefore proposes how it can be taken into consideration for risk assessment. Our assessment exploits one of the largest sample of financial institutions, which includes more than 11,000 institutions, both listed and non-listed, representing more than 180 countries over the period 2005-2014. This is the result of an intensive merge and match of data from many different sources aimed to create the most comprehensive state-of-the-art global banking set. We propose a novel indirect clustering approach in finance for business models classification able to disentangle the multi-dimensionality problem of peer group detection (groups of institutions adopting the same business model), which is compared with mainstream classification methods. We find seven specific peer groups that are consistent with the three classic business model categories discussed in previous works, such as the deposit-oriented, the wholesale-oriented and the investment-oriented model. In order to test the impact of business model to the resilience of distress, we built a comprehensive list of global distress events by combining bankruptcies, liquidations, defaults, distressed mergers, and public bailouts. We regress these events against financial statement ratios (i.e. proxies for CAMELS) and controlling for macro and sectoral effects using a rare-logit model. We observe that both wholesale-oriented and deposit-oriented models are prone to the risk of distress. However, we empirically discover different driving forces contributing to that level of instability, discouraging the one-rule-fits-all approach in favour of a more appropriate targeted intervention coherent with the institution true banking business models. Another interesting finding is the impact of business model volatility at the institution level measured by the number of switches to a different business model prior to the financial crisis. We provide supporting evidence on a consistent portion of distresses coming from institutions that migrated quite often between peer groups prior to the crisis. We confirm that the higher the volatility of business model adopted, the higher the likelihood of distress. The evolution of business models over time and the geographic composition of peer groups isalso discussed.