Rate-distortion is the fundamental trade-off at the heart of every lossy compression system: how much data (rate) do you spend to achieve how little quality loss (distortion)? Spend more bits — bigger file, less distortion. Spend fewer bits — smaller file, more distortion. The whole field of video compression is essentially the study of how to move that trade-off curve as far down-and-to-the-left as possible — less distortion at the same rate, or less rate at the same distortion.
The trade-off shows up everywhere in encoding decisions. Rate-distortion optimisation (rdo) is the encoder's algorithm for picking which coding mode to use for each block by minimising a combined cost J = D + λ·R. Rate-distortion curves plot quality (PSNR, SSIM, VMAF — any distortion metric) against bitrate, letting you compare two codecs visually: the curve closer to the top-left corner is more efficient. BD-rate (Bjøntegaard Delta-rate) summarises the difference between two rate-distortion curves as a single percentage — "AV1 has -30 % BD-rate vs HEVC" means AV1's curve is shifted leftward enough that, on average, it needs 30 % less bitrate for the same quality.
For a product team, rate-distortion is the language industry papers and vendor benchmarks use to describe codec quality. When evaluating a new codec or encoder, the question is never "is it better?" but "where is its rate-distortion curve relative to ours?". The honest answer often involves showing curves for multiple content types — sports, drama, animation, screen recording — because rate-distortion behaviour varies dramatically by content. A codec that wins on natural-motion video might lose on animation. The deliverable from any serious codec evaluation should be rate-distortion plots, not single numbers.
