Résumé : In the present article, we analyze the transaction fees market on smart contracts-enabling blockchains. On such systems, as opposed to traditional on-premise and cloud computing solutions, users are effectively competing for computational resources through an auction for priority. This paper proposes a way to estimate the bid one has to offer to have a transaction included in the next block. This method outperforms naive bidding (bidding the optimal value of the last block) if the user is realistically "impatient" to have a transaction processed. It also shows that users collectively spend several million of dollar every years for transaction fees that could be avoided without degrading the service received. This is this "waste" we seek to reduce throughour forecasting method.