A data value that falls far outside the expected range, which may indicate an error or a genuinely unusual observation.
Outliers stand out from the rest of the data. They can result from data entry errors, measurement problems, or system glitches — but they can also be real and meaningful. A property assessed at $50 million in a neighbourhood where values average $400,000 might be an error, or it might be a legitimate commercial property.
The key question with any outlier is: is this a data quality problem or a real phenomenon? Investigating outliers is an important part of data profiling and quality assurance.
Example: A dataset of daily temperature readings for a weather station shows a value of -999°C on one day. This is clearly a sentinel value used to indicate missing data, not an actual temperature. Treating it as a real value would corrupt any analysis.