What we call a single row of data. For example, if you have a list of people who have donated to a nonprofit before all lined up in a spreadsheet, one row = one record. What exactly the record contains depends on how many features it has. It could just be a name, if that’s all you collected. It could be name, address, phone number, pronouns, and 10 other things. The thing that matters here though is that the record is just one line of data. The algorithm learns from records while in training.