Peer-to-Peer (P2P) FinTech platforms allow cost reduction and service improvement in credit lending. However, these improvements may come at the price of a worse credit risk measurement, and this can hamper lenders and endanger the stability of a financial system. We approach the problem of credit risk for Peer-to-Peer (P2P) systems by presenting a latent factor-based classification technique to divide the population into major network communities in order to estimate a more efficient logistic model. Given a number of attributes that capture firm performances in a financial system, we adopt a latent position model which allow us to distinguish between communities of connected and not-connected firms based on the spatial position of the latent factors. We show through empirical illustration that incorporating the latent factor-based classification of firms is particularly suitable as it improves the predictive performance of P2P scoring models.