Click-through rate (CTR) measures how often a presented recommendation earns a click. In a movie recommendation page, it is the percentage of users who click at least one recommendation after the page loads. CTR connects technical performance to user engagement, but it should be read alongside downstream consumption and other product outcomes. section 4.4.3
Use CTR as one part of an evaluation anatomy:
- Pair it with system metrics such as p50, p95, and p99 page-load time to connect latency with behavior. section 4.4.3
- Compare it with bounce rate, abandonment rate, and downstream consumption rather than treating a click as the final outcome. section 4.4.3
- Segment experiments by device type because latency and engagement effects can differ across desktop, iOS, and Android. section 4.4.3
- Revalidate CTR as a proxy when models or products change; generated explanations can make clicks persuasive without making recommendations satisfying. section 8.5.2
Latency degradation in movie recommendations
A mobile experiment compared a control against five added-latency variants. CTR fell from 23.5% with no added latency to 8.7% with 500 milliseconds added, while average page-load time rose from 1050 to 1552 milliseconds and bounce rate increased from 27.8% to 56.9%. The result links slower delivery with weaker immediate engagement, but downstream viewing is still needed to assess recommendation value. section 4.4.3




