Causal Data Science for Business Analytics

Master Course

Author

Christoph Ihl

Published

July 15, 2026

0.1 Overview

Most managerial decision problems require answers to causal questions such as “what happens to Y if we do X?” or “was it X that caused Y to change?” This course teaches students to distinguish causal relationships from mere correlations using modern data science approaches.

We cover confounding factors, selective measurement in data collection, and extrapolating causal knowledge across different business contexts. The course progresses from foundational causal language (DAGs, potential outcomes) through experimental and quasi-experimental designs to modern causal machine learning methods and panel data techniques.

Three pillars distinguish this course:

  • Graphical and formal causal reasoning. Students learn to formalize causal intuitions using directed acyclic graphs (DAGs) and the potential outcomes framework — moving beyond “correlation is not causation” to precise identification arguments.

  • A complete quasi-experimental toolkit. The course covers matching, instrumental variables, regression discontinuity, difference-in-differences, and synthetic controls — the core designs that allow causal claims from observational data.

  • Causal machine learning. Double machine learning, causal forests, and meta-learners bridge traditional econometrics and modern ML — enabling high-dimensional control, heterogeneous treatment effects, and data-driven targeting.

Each week combines a lecture introducing concepts and methods with a problem-based learning (PBL) session where students apply the methods to real and simulated datasets using R or Python.

0.2 Learning Objectives

Upon completion of this course, students will be able to:

  1. Distinguish between “correlation” and “causation”
  2. Recognize the shortcomings of purely correlation-based approaches
  3. Develop causal knowledge for data-driven decision-making
  4. Formalize causal intuitions using causal inference language
  5. Derive testable causal hypotheses
  6. Discuss state-of-the-art causal data science tools and algorithms
  7. Conduct independent causal data analyses

0.3 Target Audience

  • Master students in Data Science and IWI programs
  • Master students with strong interest in causal aspects of modern business analytics

Credits: 6 ECTS

0.4 Time & Location

Component Time Location
Causal Data Science (Lecture) Monday, 11:30–13:00 Building D, Room D-1.023
Business Analytics with Causal Data Science (PBL) Tuesday, 15:00–17:15 Building D, Room D-0.013

0.5 Schedule

# Date Topic
01 2026-04-13 Introduction to Causal Inference
02 2026-04-27 Graphical Causal Models
03 2026-04-28 Causal Graphs and Discovery
04 2026-05-04 Experiments & Linear Regression
05 2026-05-18 Online A/B Testing
06 2026-06-01 Observed Confounding
07 2026-06-08 Instrumental Variable Designs
08 2026-06-15 Double Machine Learning
09 2026-06-22 Treatment Effect Heterogeneity
10 2026-06-29 Difference-in-Differences
11 2026-07-06 Synthetic Controls
12 2026-07-13 Counterfactual Imputation

The course follows a methods-building arc across five parts. Each week combines a lecture introducing causal inference methods with a PBL session where students apply the methods to data.

Part Sessions Focus Assignment
I. Causal Language S01–S03 DAGs, potential outcomes, causal discovery Assignment I
II. Experiments S04–S05 RCTs, A/B testing, interference Assignment II
III. (Un-)Observed Confounding S06–S07 Matching, IV, RDD, sensitivity analysis Assignment III
IV. Causal Machine Learning S08–S09 Double ML, causal forests, meta-learners, policy learning Assignment IV
V. Causal Panel Data S10–S12 DiD, synthetic controls, counterfactual imputation Assignment V

0.6 Grading

Assessment combines take-home analytical reports with in-class quizzes to ensure genuine understanding. Assignments may be submitted in R or Python.

Component Weight What it captures
5 Assignments (report + quiz each) 5 × 20% Causal reasoning, method implementation, interpretation

0.6.1 Report

Each assignment is a take-home analysis submitted as a rendered Quarto document with integrated code, output, and interpretation. Reports are graded on methodology, interpretation, and code quality.

0.6.2 Quiz (Gate)

A 15-minute closed-book paper quiz at the Monday lecture following each of the first four submission deadlines (Assignments I–IV). Each quiz consists of 2 multiple-choice questions (5 points each) and 2 short-answer questions (15 points each), totaling 40 points.

Exception (Assignment V / Part V): Assignment V has no quiz. As the Part-V capstone — synthesising difference-in-differences, synthetic control, and counterfactual imputation on a single dataset with a graded assumption ledger and board memo — the assignment is itself the comprehension gate for the final part, so no separate quiz is administered. The quiz-gate rules below (50% threshold, 60% cap, oral-defense trigger) therefore apply to Assignments I–IV only.

For Assignments I–IV, you must score at least 50% on the quiz to receive your full report grade. Below 50%, the assignment grade is capped at 60%.

0.6.3 Oral Defense

Students whose quiz performance is significantly below their report quality will be asked to do a short oral defense during office hours to keep their report grade.

0.6.4 Deadlines

Assignment Submission (Sun 23:59) Quiz
Assignment I May 17 Mon May 18 (first 15 min, S05 lecture)
Assignment II May 31 Mon Jun 1 (first 15 min, S06 lecture)
Assignment III Jun 14 Mon Jun 15 (first 15 min, S08 lecture)
Assignment IV Jun 28 Mon Jun 29 (first 15 min, S10 lecture)
Assignment V Aug 2 — (no quiz; A5 is itself the Part-V gate)

0.6.5 Oral Defense Timeline

  • Flagged students notified by email by Friday 23:59 of the quiz week with an office hours slot.
  • Defense completed within 2 weeks of the quiz — before the next assignment cycle.

0.6.6 Late Submission Policy

  • Up to 24 hours late: -20% penalty on the report grade.
  • More than 24 hours late: report not accepted (assignment grade = 0).
  • Extensions: must be requested via email before the original deadline with a valid reason (e.g., medical certificate).

0.6.7 Quiz Absence Policy

  • Excused absence (medical certificate submitted within 3 days): makeup quiz during office hours within one week of return.
  • Unexcused absence: quiz score = 0 — gate fails, assignment grade capped at 60%.

0.7 Registration

Register for the entire module at e-learning.tuhh.de/studip.

0.8 Contact

Christoph Ihl (christoph.ihl@tuhh.de)