Most data arrives as a CSV export. read_csv() from readr reads it into a tibble and guesses each column's type from the first rows: numbers, dates written as 2026-03-02, TRUE and FALSE, or text. It prints the types it chose, so read that message every time.
At work you pass a file path, read_csv("exports/shipments.csv"). Here the exports are already in the session as text, so you wrap the text in I() to tell readr it is the data itself: read_csv(I(raw)).
Three arguments fix most bad imports. na = c("", "NA", "n/a") lists what counts as missing; anything else stays text and turns a number column into a text column. col_types = cols(shipped = col_date("%m/%d/%Y")) overrides a guess, for dates written month first. skip = 2 jumps over notes at the top of the file.
Read with readr's guesses, say what missing looks like, skip notes and fix one column's type.