In this talk, we will discuss recent work on managing tradeoffs between efficiency and effectiveness in modern multi-stage ranking architectures which are comprised of a candidate generation stage followed by one or more reranking stages. In such an architecture, the quality of the final ranked list is often sensitive to the quality of initial candidate pool. We will briefly discuss a few recent related papers from my group, and then discuss future directions. First, we will explore dynamic cutoff prediction in early stage retrieval using query difficulty pre-retrieval features. We will then turn our attention to efficiency and effectiveness trade-offs in the later stage cascaded learning-to-rank algorithms. Specifically, we reexamine the importance of tightly integrating feature costs into multi-stage learning-to-rank (LTR) IR systems, and we present a novel approach to optimizing cascaded ranking models which can directly leverage a variety of different state-of-the-art LTR rankers such as Lamb-daMART and Gradient Boosted Decision Trees. Finally, we discuss interesting future research directions in multi-stage retrieval systems as modern retrieval tasks continue to evolve towards more complex interactive search systems.