We study a two-echelon supply chain with first order autoregressive demand and unit replenishment lead-times. Each echelon of the supply chain uses conditional expectation to generate Minimum Mean Squared Error forecasts. Both echelons use these forecasts inside the ‘Order-Up-To’ policy to generate replenishment orders. We investigate three different scenarios. The first is when each echelon aims to minimise their own local inventory holding and backlog costs. The second scenario is concerned with an altruistic retailer who is willing and able to sacrifice some of his own performance for the benefit of the total supply chain. The retailer does this by smoothing the demand placed on the manufacturer by using a matched proportional controller in the inventory and Work-In-Progress feedback loops. The third scenario is concerned with an altruistic retailer with two, unmatched controllers. The matched controller case outperforms the traditional case by 14.1%; the unmatched controller case outperforms the matched controller case by 4.9%.