Skip to content

Business Problem — COVID-19 Data Analysis

Context

You're a data analyst at a public health think tank producing weekly briefings for policymakers and journalists. The briefings must be accurate, comparable across countries, and free of the distortions that plagued much early pandemic reporting (raw case counts that ignored population, testing differences, and reporting lags).

The challenge: raw COVID numbers are misleading. The US having more total cases than Iceland says nothing about which country handled the pandemic better — the US has 1,000× the population. Comparable, honest analysis requires per-capita normalisation, rolling averages, and careful caveating.


Analytical Questions to Answer

Q1 — Global Progression

  • How did global cases and deaths evolve over time?
  • Can we identify the distinct "waves" of the pandemic?
  • When did each wave peak globally?

Q2 — Fair Country Comparison

  • Which countries had the highest cases per capita (not in absolute terms)?
  • Which had the highest deaths per capita?
  • How does the case fatality rate (deaths / cases) vary across countries?

Q3 — Wave Analysis

  • For a selected set of countries, when did each wave hit?
  • How did the severity of waves change over time?
  • Did later waves have lower fatality rates (better treatment, vaccines)?

Q4 — Vaccination Impact

  • How did vaccination rollout progress across countries?
  • Is there a visible relationship between vaccination rate and death rate in later waves?

Q5 — Communication

  • How do we present this data without sensationalising or downplaying?
  • What caveats must accompany every comparison?

Why This Project Matters for Analysts

COVID-19 data is a masterclass in the things that separate good analysts from dangerous ones:

Trap Lesson
Raw counts ignore population Always normalise per capita for comparison
Cumulative ≠ daily Know whether your metric is a running total or a daily value
Reporting noise (weekend dips) Use 7-day rolling averages
Testing differences Cases depend on testing — caveat heavily
Cherry-picked date ranges Show full context, not convenient windows
Correlation ≠ causation Vaccination correlations need careful framing

Stakeholders

Stakeholder Needs
Policymakers Accurate, comparable trends to inform decisions
Journalists Clear visuals with honest caveats
The Public Trustworthy information, not fear or false reassurance

← Project Overview · Next: Dataset Guide →