Per-title encoding tunes the encoding ladder to each individual piece of content instead of applying one fixed ladder to everything. A simple animation and a grainy, fast-moving action film have very different complexity, so the bitrate needed to look good differs enormously. A fixed ladder over-spends bits on easy content and under-delivers quality on hard content; per-title encoding fixes both.
In practice the encoder analyzes each title (often with multiple test encodes or a complexity model) and chooses the optimal resolution-and-bitrate points for that specific content — fewer rungs, lower bitrates, or different resolution crossover points. The payoff is real: the same perceived quality at materially lower average bitrate, which directly cuts the dominant egress cost, multiplied across a large catalog and millions of streams.
Context-aware encoding extends the idea further by also factoring in the viewing device and network, and modern pipelines lean on perceptual metrics like VMAF to decide where quality is "good enough."
