1. By Role or Specialization (The “Who does what?”)
This is probably the most practically useful classification for an aspiring data scientist.
The practice of data science has diversified into multiple specialities, each being loosely defined.
| Specialization | Primary Focus | Key Skills | Common Job Titles |
|---|---|---|---|
| Data Analyst | Finding insights from past data to answer business questions. | SQL, Excel, BI Tools (Tableau, Power BI), Statistics, Visualization. | Business Analyst, Reporting Analyst, BI Analyst |
| Data Scientist | Building predictive models and machine learning systems to solve complex problems. | Python/R, SQL, Machine Learning, Statistics, Deep Learning, Big Data Tools. | Machine Learning Scientist, Data Scientist |
| Data Engineer | Building and maintaining the data infrastructure (pipelines, warehouses) that enables analysis. | SQL, Python, Java/Scala, Big Data Tech (Spark, Hadoop, Kafka), Cloud Platforms (AWS, GCP, Azure), ETL. | Data Architect, Big Data Engineer |
| Machine Learning Engineer | Taking models from prototype to production; focuses on scalability, efficiency, and deployment. | Python, Software Engineering, ML Frameworks (TensorFlow, PyTorch), MLOps, Cloud, Docker/Kubernetes. | ML Ops Engineer, AI Engineer |
| Statistician | Designing experiments and using rigorous statistical methods for inference and understanding causality. | R/Python, Experimental Design, Statistical Theory, Hypothesis Testing. | Research Scientist, Biostatistician |
| Business Intelligence (BI) Developer | Creating easy-to-use dashboards and reports for business users to monitor KPIs. | SQL, BI Tools (Tableau, Power BI, Looker), Data Warehousing. | BI Analyst, Reporting Specialist |
Analogy: If data science were a restaurant:
Data Engineer – The kitchen manager and supply chain. They keep the ingredients (data) in stock and ensure that the quality is consistent.
Data Scientist – The chef. They make the recipes out of ingredients.
ML Engineer – The kitchen tech who takes the chef’s recipe and makes it production-ready for the whole restaurant chain.
Data Analyst – The food critic and menu planner. They analyze which dishes are popular and plan the menu.
BI Developer – The menu designer. They design the dashboards (menu) on which the waiter will point the customers’ attention to
2. By Data Type (The “What data is being used?”）
Depending on the nature of the data itself, there are several types of data science:
Structured Data: Analysis of tabular data found in databases (typically SQL tables or Excel spreadsheets).
Unstructured Data: Analysis of non-structured data (much more complex).
Natural Language Processing : Understanding and extracting meaning from text data (reviews, documents, social media).
Computer Vision: Analysis of visuals (images and videos).
Audio Processing: Analysis of sounds and voices.
Geospatial Data: All data with a geographical component (on maps, coordinates). Used in logistics, real estate, and environment.
Graph Data: Data involving graphs and networks (social networks, PNRs, fraud detection).
Time Series: A sequence of data points, typically measurements made over time (stock prices, sensors, weather).
3. By Methodology or Approach (The “How is it done?”)
These are types related to the purpose of the analysis or modeling
Descriptive Analytics: “What happened?” The most basic type, which only uses historical data to identify trends or patterns (e.g., “Sales in Q3 were down 10%”).
Diagnostic Analytics: “Why did it happen?” Looks at reasons for a phenomenon or some peculiarities that happened (e.g., sales drop was caused by a competitor’s product launch).
Predictive Analytics: “What is likely to happen?” Uses statistical models and machine learning to predict future occurrences (e.g., customer churn in the next quarter is expected to rise by 15%).
Prescriptive Analytics: “What should we do?” The most sophisticated type, which proposes courses of action to either mitigate or capitalize on expected outcomes (e.g., it would be beneficial to provide a 20% discount to 500 of these customers to reduce the estimated churn).
Causal Inference: “Did X cause Y?” Describes cause-effect relationships, which go beyond mere correlation, often employing controlled experimentation (A/B testing) (e.g., was the website redesign responsible for the rise in sign-ups?).
Summary
When someone asks about types of data science, they most likely mean the types in the specializations and roles. However, any real-world data science project will have elements of all three described above.
For instance, a Data Scientist (Role) could build a Computer Vision model (Data Type) to Predict (Methodology) defects on an assembly line.
