Traditional quality assurance methods for large-scale video distribution networks operate independently at different points along the video delivery chain, reporting partial and incoherent measurements, leading to poor and fragmented understanding about how multiple stages of quality degradations affect the final quality-of-experience (QoE) of end users. We propose a framework that uses a unified end-to-end solution to produce consistent QoE scores at all points along the delivery chain under the same evaluation criterion. The novel solution produces a clear and complete picture instantaneously about how video QoE degrades over the network, allows immediate issue identification, localization and resolution, enables quality and resource optimization, and provides reliable predictive metrics for long-term strategic resource and infrastructure allocations. The main challenge in the implementation of the solution is to create a unified QoE metric that not only accurately predicts human perceptual QoE, but is also lightweight and versatile, readily plugged into multiple points in the video delivery chain. The QoE metric should produce real-time QoE scores across a wide range of bitrates, resolutions, frame rates and dynamic ranges, and combine presentation picture quality with the perceptual impact of video freezing and adaptive streaming events. We show that the SSIMPLUS metric offers the best promise to meet all the challenging demands.