Data science is the discipline of pulling usable knowledge out of data using statistics, programming and domain understanding together. None of the three is optional. Statistics without domain knowledge produces technically correct answers to the wrong question, and domain knowledge without method produces confident guesses.
The work runs across a familiar arc. Framing the question, collecting and cleaning data, exploratory analysis, modelling, evaluation, and communicating the result. In real projects the cleaning stage takes the largest share of the time, which surprises people who expect modelling to dominate.
It overlaps with neighbouring roles without being identical to them. Data engineering owns the pipelines and storage, machine learning engineering owns getting models into production, and analytics owns describing what happened. Data science sits across those and leans toward answering why and what next.
A concrete case: an ecommerce team asks why repeat purchase rates fell. A data scientist joins order, campaign and site behaviour data, and finds the decline concentrated in customers acquired through one channel during one quarter. The answer is not a model, it is a segmented comparison that survives scrutiny.
The hardest part is usually the question rather than the method. A precisely stated question with an agreed measure of success can be answered with simple tools, while a vague one resists the most sophisticated model.
