Key Takeaways
- Data science is a communication job as much as a technical one.
- SQL, Python, statistics and one visualisation tool are non-negotiable.
- Career progression flows Analyst → Data Scientist → Senior DS → Applied ML.
- Portfolio projects matter more than certificates from every angle.
- The interview loop tests case reasoning, coding, statistics and product sense.
The Four Pillars
| Pillar | Concrete Skills |
|---|---|
| Programming | Python, pandas, numpy, notebooks, git |
| Data | SQL, joins, window functions, warehouse basics |
| Statistics | Distributions, hypothesis testing, regression, A/B testing |
| Communication | Framing, visualisation, executive summaries |
If any pillar is weak, the whole role is weak. Do not skip statistics because it is hard.
Twelve-Month Plan
- Months 1-2 — Foundations. Python, pandas, plotting, basic SQL, descriptive statistics.
- Months 3-4 — Intermediate. Advanced SQL, probability, hypothesis testing, sklearn basics.
- Months 5-6 — Project 1. End-to-end analysis on a public dataset — Kaggle or open government data — with a written report.
- Months 7-8 — Modelling. Regression, tree-based methods, cross-validation, evaluation, feature engineering.
- Months 9-10 — Project 2. A predictive model deployed as a small dashboard or API.
- Months 11-12 — Interview prep. Case studies, SQL drills, product-sense practice.
Portfolio That Convinces Hiring Managers
- One deep analysis with strong writing.
- One predictive model with an honest evaluation section.
- One dashboard non-technical viewers can use.
- One small A/B-testing simulation or causal-inference case.
Publish everything on GitHub with clean READMEs. Recruiters skim READMEs, not notebooks.
The Interview Loop
- SQL screen — window functions, joins, aggregation.
- Python / pandas screen — data cleaning and feature construction.
- Statistics — sampling, tests, confidence intervals, A/B tests.
- Case study — how would you measure, model or improve X.
- Behavioural — projects, trade-offs, stakeholder collaboration.
Common Mistakes
- Doing courses instead of projects.
- Learning frameworks before fundamentals.
- Choosing overly novel datasets no interviewer can evaluate.
- Ignoring SQL because it feels basic — every serious loop tests it.
- Talking about accuracy without talking about the business metric.
Final Summary
Data science rewards patience. Learn the fundamentals thoroughly, ship a small number of high-quality projects and be honest about what your work does and does not prove.