Rakidzich Research

How this analysis was produced

Back to the findings

Corpus
15 publicly published student interviews
Selection rule
Every student interview linked from the channel's published success-stories page was included. No interview was excluded on the basis of what it contained. The corpus is therefore the set the channel chose to publish, and is not a random or representative sample of students.
Transcription
Machine-generated captions fetched from the published videos. No human transcription and no model transcription were used.
Claim extraction
Deterministic marker windows. A window opens around a currency, percentage, unit-count or cadence marker, overlapping windows merge, and the merged span is capped. No model participates at any point.
Numeric normalisation
Digit, k-suffix, space-grouped and spoken-word forms are parsed to integers. A digit amount absorbs a following scale word, so '$42 million' is one value.
Aggregation
Deterministic. Rerunning on the unchanged corpus reproduces every figure.
Speaker attribution
Not established. Utterances carrying explicit host-framing phrases are excluded; the remainder is recorded as unestablished, never as the student.
Attribution recall
Interviewer speech is detected by explicit host-framing phrases only. That test is precise but its recall is unmeasured, so an excerpt marked unestablished is not thereby the student's: it is an utterance in which no host phrase was found.
Quote handling
Excerpts are machine-generated captions. They carry no punctuation in the source and none has been added. Vocal fillers (um, uh) were removed and the first word was capitalised; every other word is exactly as the captions recorded it. An ellipsis marks where an excerpt was cut out of a longer stream.
Interview dates
Not available. The transcript source returned no publication date for any of the 15 interviews, so this analysis claims no time period for the corpus and the dataset declares no temporalCoverage.
Revenue verification
None.
Property verification
None.
Analysis date
2026-09-06

What this data does not show