Data Engineering vs Data Science
PedalCity wants to predict how many bikes each station will need tomorrow morning, so vans can move bikes overnight. That one project needs several kinds of work, and people often blur them together. Knowing who builds what stops a team from asking a statistician to fix a broken pipeline, or an engineer to invent a forecasting method.
The rule
Engineers make data trustworthy; scientists make it answer "why" and "what next".
The roles overlap. Both write SQL and Python, and both care about data quality. The difference is what they hand over at the end:
| Role | Main question | Typical deliverable |
|---|---|---|
| Data engineer | Is the right data there, correct and on time? | Pipelines, tables, data models |
| Data analyst | What happened? | Reports, dashboards, answers to business questions |
| Data scientist | Why did it happen, and what will happen next? | Models, experiments, forecasts |
For the bike forecast, the data scientist chooses the model and tests how accurate it is. The data engineer delivers years of clean hourly dock counts and ride history, and then feeds the model fresh inputs every night. An analyst might report last month's shortages that motivated the project.
A useful fact to remember: most of a model's quality comes from its input data. A brilliant model trained on data with gaps and duplicates makes confident, wrong predictions.
The shortcut
Ask what the deliverable is. A pipeline or a table points to the engineer. A model, experiment or explanation points to the scientist. A report on what already happened points to the analyst.
"Collect hourly dock counts since 2023 into one table"
The deliverable is a table fed by a pipeline, so this is data engineering.
Drills
Answer fast — trust the rule.
"Build a job that lands hourly dock counts in a table every hour." Whose core work is this?
"Build a model that predicts how many bikes each station needs tomorrow." Whose core work is this?
The forecast gets worse because dock readings arrive with gaps. What should be fixed first?
Which statement is true?
Remember
- Engineers make data trustworthy; scientists make it answer "why" and "what next".
- Ask what the deliverable is: pipeline or table, model or experiment, report on the past.
- A great model cannot fix bad input data.