The effort required to perform data wrangling operations, also known as data munging, is an understated aspect to all data science activities. Online courses or web-based examples generally provide pre-cleansed datasets for end users. This may give the impression that real-world data is similar to that used for data mining exercises and/or courses. In fact, real-world data is seldom, if ever, anywhere close to the pristine datasets depicted in such courses.
Real-world data will very likely not be in the format you need for your machine learning activities, may contain inaccurate or missing data, have mixed data types in the same column (for example, numbers and characters in the price column), and pose a host of other challenges that few of us are prepared for at the onset.