Data Analytics & Analytics Engineering

Take raw, messy business data through the full analytics chain, from SQL and Python to dashboards and a recommendation you can defend.

6 Weeks3 Sessions/WeekIn PersonBeginner-friendly

SQLData ModellingDashboards

Share this course

Overview Curriculum Capstone Assessment Enquire

What you'll be able to do

Graduates can take raw, messy, multi-table business data through the full analytics chain to a decision.

Turn a vague business question into a defined, answerable metric
Model and query relational data in SQL, including cohort, funnel and window analysis
Clean and analyse messy real-world data in Python
Build tested, documented, version-controlled transformation pipelines
Build dashboards that answer stated business questions
Quantify uncertainty and design a valid A/B test
Present a recommendation to a non-technical decision-maker and defend it

Who it's for

Commerce, economics, statistics, management and social science graduates
Working professionals in finance, operations, marketing or research
Analysts who have reached the limits of spreadsheet work
Anyone entering a data career without a computer science background

Prerequisites

Spreadsheet familiarity
Comfort working with numbers
No programming experience required

All students complete the two-session Engineering Onboarding module before Module 1.

Tools and technologies

PostgreSQL and SQLPython with pandasdbt-style transformation modellingPower BI (primary)Metabase (open-source alternative)Git and GitHubscipystatsmodelsAn LLM assistant used under a verification discipline

Target roles

Data AnalystBusiness or MIS AnalystBI AnalystJunior Analytics EngineerReporting Analyst

Course curriculum

6 modules · 5-6 weeks

Concepts: what an analyst is paid to produce; converting a business question into a specific answerable one; the stakeholder interview; metric definition and why definitions cause more disagreement than data; dimensions against measures; grain; SQL covering selection, filtering, sorting, NULL semantics and the errors they cause, aggregation, grouping, filtering on aggregates, all four join types, set operations, conditional expressions, and date and string handling; reading an unfamiliar schema; data dictionaries.
Lab: given an undocumented multi-table business database covering customers, orders, payments, support tickets and staff, reverse-engineer and document the schema; complete thirty progressively harder queries; convert five vaguely worded stakeholder requests into precise metric definitions and then into SQL.
Project: a documented data dictionary and metric catalogue for the working database.

Concepts: subqueries and common table expressions, and why the latter make analysis reviewable; window functions covering ranking, running totals, moving averages, period-over-period comparison and first and last value; self-joins; pivoting; cohort and retention analysis; funnel analysis and drop-off; time-series aggregation and calendar tables; slowly changing data; query performance covering indexes and execution plans; SQL style and reviewability; verifying that a query is correct rather than assuming it.
Lab: build cohort retention, a conversion funnel with drop-off rates, month-over-month growth and an RFM segmentation entirely in SQL; reduce a forty-second query to under two seconds; locate the fault in a query returning plausible but incorrect numbers caused by row fanout from a join.
Project: Mini-project 1: Business Performance Analysis. A set of reviewed, documented SQL analyses answering six business questions, with findings written for a manager.

Concepts: Python essentials for analysis; pandas covering loading from files, databases and APIs, selection, filtering, grouping, merging, reshaping, and the performance cost of row-wise operations; real-world cleaning covering inconsistent categories, duplicate detection and entity resolution, date and encoding problems, mixed types, outliers, and missing-data mechanisms with defensible handling strategies; validating data against expectations; multi-file and multi-sheet ingestion; automating a recurring manual report; visualisation for exploration against visualisation for presentation; reproducible analysis structure and version control for analytical work.
Lab: clean a poor-quality spreadsheet export containing merged headers, mixed date formats, trailing whitespace, duplicate entities and numeric values stored as text, producing an analysis-ready table with a documented cleaning log; automate a recurring manual report and measure the time saved; profile a dataset and produce a data-quality report.
Project: a reproducible ingestion and cleaning script in version control, running from raw source to clean table with one command.

Concepts: why dashboards built directly on production tables fail; the layered warehouse model of raw, staging, intermediate and mart tables; dimensional modelling covering facts, dimensions, star schemas, declared grain and surrogate keys; transformation as version-controlled, tested code, including modular models, references, seeds and incremental logic; data tests covering uniqueness, null constraints, referential integrity, accepted values and freshness; documentation and lineage generated from code; orchestration and scheduling, idempotency and backfills; pipeline failure handling and alerting; the analytics engineering career step.
Lab: build the full layered model from the raw database up to two mart tables; add data tests, then corrupt the source and confirm the tests catch it; generate documentation and a lineage graph; schedule the pipeline and configure failure alerting; review a classmate's models for grain errors.
Project: Mini-project 2: a tested, documented, scheduled analytics pipeline in a repository, producing the mart tables used by the dashboards.

Concepts: dashboard design starting from the decision it supports; audience-appropriate design across executive, operational and analytical use; chart selection and the small set that reliably works; presentation honesty covering axis truncation, dual axes, selective date ranges and misleading aggregation; colour, accessibility and layout; interactivity, filters and drill-down; dashboard performance at real data volume; refresh, row-level security and access control; a consistent metric layer so two dashboards cannot disagree; data storytelling that leads with the finding, quantifies the impact, states confidence, recommends an action and names what would change the conclusion; the one-page written summary; presenting to an audience that will challenge the numbers.
Lab: build an executive dashboard and an operational dashboard from the same mart tables and explain why they differ; critique three misleading published charts and rebuild them honestly; present findings to a panel role-playing a sceptical management team.
Project: a dashboard suite on the Module 4 pipeline, with a one-page written summary.

Concepts: descriptive statistics and distribution shape; sampling and sampling error; confidence intervals expressed in plain language; practical hypothesis testing and the correct interpretation of a p-value; A/B testing covering hypothesis, metric, sample size, duration, guardrail metrics, the cost of early peeking and the common invalidating errors; correlation, causation and confounding; segmentation and Simpson's paradox; forecasting basics and seasonality; effect size against statistical significance; LLM-assisted analysis covering query and code generation, explaining unfamiliar queries, and classifying qualitative text such as support tickets and survey responses at scale, alongside the verification discipline that the analyst owns every number presented.
Lab: design an A/B test for a product question including sample size and guardrails, then analyse supplied results and write the recommendation; identify a Simpson's paradox in the working dataset; generate three complex queries with an LLM, verify each against a known-correct answer and document the errors; classify 500 free-text support tickets and validate against a hand-labelled sample.

Capstone project

A full analytics project addressing a real organisational problem. Options include customer churn and product holding analysis for a bank, patient flow and appointment no-show analysis for a hospital, retention and margin analysis for an e-commerce business, or programme outcome analysis for a development organisation.