
TV Series · 2022 · TV Series · 2020s
Severance
Scored from 1,257 calibrated IMDb reviewers, each re-centered on their own rating baseline before the title is ranked. Cohort: TV Series · 2020s (3,070 peers).
Summary
At a mysterious corporation called Lumon Industries, employees undergo the Severance procedure, a surgical process that divides their memories—creating separate 'innie' and 'outie' consciousnesses with no recollection of each other. Mark Scout and his colleagues uncover disturbing truths about their employer and the nature of the procedure.
Severance is a 2022 science fiction, thriller and drama television series created by Dan Erickson. Its certificate is TV-MA. Its country of origin is listed as the United States. Episodes run about 50m. Mark Ruffalo, Adam Scott and Patricia Arquette head the billed cast.
The cohort it is ranked inside is 2020s television series — 3,070 of them. The calibrated figure is built from 1,257 reviewers whose rating history varies enough to place their own average. A broader pool of 1,337 reviewers with two or more reviews feeds the raw-average track instead. Rotten Tomatoes carries a separate audience for the same title, 718 of whom clear the calibration test.
How the score was built
We compare the reviewers of this title against every other title those same people reviewed. Each review is scored relative to the reviewer's own baseline — a 7/10 from a hard grader sits well above their average, while a 7/10 from a generous grader sits below theirs. Every reviewer then counts the same, so the result is a calibrated consensus rather than a raw average.
- Global percentile
- Where Severance lands against every title we rank.
- In-cohort percentile
- The same comparison narrowed to its cohort — titles of the same format released in the same decade. A 1970s film is judged against 1970s films, not against last year's streaming series. Ranked against TV Series · 2020s (3,070 titles).
- AI-adjusted percentile
- The same reviewer-normalized score, corrected for how much confidence the sample supports: thin samples are pulled toward the corpus average. Each title keeps the share n / (n + 53) of its measured distance from the average, where n is its calibrated-reviewer count, and the result is ranked on the same scale as the global percentile.53 is the median sample size across everything we rank. Here n is 1,257.
Cohort: TV Series · 2020s



