This paper examines the service design challenges faced by an on-demand delivery platform serving consumers who are not only averse to congestion but also exhibit asymmetric preferences for losses and gains. The platform needs to choose a delivery time mode between time point (TP) in which the platform promises to deliver by a time point, and time window (TW) where the platform promises the delivery within a time window. Our results indicate that consumer loss aversion is a prerequisite for platforms to use TW. The platform also decides whether to offer service enhancement strategies including on-time guarantee (OTG) and fast delivery (FD). For the former, specifically, we investigate both the scenario of the platform offering on-time guarantee as an integral part of the delivery service without charging an extra fee (referred to as PG), and the scenario of consumers paying for the on-time guarantee service separately (referred to as CG). We also find that PG (CG) should be adopted when consumers have medium (high) valuations. The adoption of FD heavily depends on the scheduling cost of the platform; a high (low) cost discourages (encourages) the usage.