Skip to content

The data-science cycle in one page

Every investigation in this book — from a Grade 8 class survey to a Grade 12 client capstone — follows the same five steps, called PPDAC. The steps are the same; what you do inside each step grows every year.

        ┌────────────┐
        │  PROBLEM   │  What do we want to know, and why? Who will use the answer?
        └─────┬──────┘
              ▼
        ┌────────────┐
        │    PLAN    │  What data, from whom, how? What could go wrong? What will we compute?
        └─────┬──────┘
              ▼
        ┌────────────┐
        │    DATA    │  Collect or download; record source and licence; clean and log every step.
        └─────┬──────┘
              ▼
        ┌────────────┐
        │  ANALYSIS  │  Summarise, graph, compare, model — and check how sure we can be.
        └─────┬──────┘
              ▼
        ┌────────────┐
        │ CONCLUSION │  Claim, evidence, reasoning, limits. Communicate. New questions?
        └─────┬──────┘
              └──────────► back to PROBLEM (every good answer raises a new question)

How each step grows from Grade 8 to Grade 12

Step Grade 8 Grade 9 Grade 10 Grade 11 Grade 12
Problem statistical vs non-statistical questions questions about districts and rates questions about time and relationships questions about uncertainty and change a client's decision and success measures
Plan survey design, fair samples, consent plan card, choice of rate cleaning plan, limit values pre-registration, seeds, Git task card, data agreement
Data tallies and tidy tables spreadsheets, cleaning logs, SQL pandas, dates, missing values big data, joins, maps, licences features, encoding, leakage checks
Analysis graphs, mean/median, rates pivots, SQL, pandas, charts distributions, best-fit lines, models bootstrap, tests, regression, trends ML workflow, metrics, audits
Conclusion claim–evidence–reasoning poster poster + SQL report report or dashboard reproducible notebook memo, model card, handover

Five questions to ask at every step

  1. Who collected these data, why, and how?
  2. Who or what is missing?
  3. Is this difference or pattern bigger than chance would produce?
  4. Could something else explain it (a confounder)?
  5. What can't we say from this — and who could be harmed if we get it wrong?

These five questions appear, in growing depth, in every grade. If you remember nothing else from this book, remember them.