Reservation prices have proven effective in boosting revenue in Generalised Second Price (GSP) auctions, particularly in cost-per-click (CPC) settings. However, in domains like music streaming, where ads are consumed passively without user clicks, a cost-per-impression (CPM) model is more appropriate. Additionally, in the music streaming domain, user intent is typically unknown, unlike in sponsored search, making it essential to optimally leverage all available user and contextual information when setting prices. This paper addresses the challenge of optimising reservation prices in GSP auctions with CPM pricing, adopting a personalised approach that accounts for both user- and advertiser-specific factors. Using dataset of 100,000 auctions from a major music streaming service, we determine such optimal prices. To achieve this, we first derive the symmetric Nash equilibrium for GSP auctions in a CPM context. We then introduce a Deep Neural Network-based mixture density model that incorporates this equilibrium into its loss. This model captures advertisers’ diverse preferences by learning directly from bidding data. We show how this approach enables the computation of personalised prices for both users and advertisers, boosting auction revenue by an average of +4% across ten markets. Our study further highlights the impact of market competitiveness and advertiser preference heterogeneity on revenue gains, showing that personalised pricing greatly enhances auction performance.