Group Work Brief

Summary

In groups of 2–3, you will develop and analyze a realistic business analytics case grounded in reputable academic or industry sources. You will formulate a concrete business problem, generate a reproducible synthetic dataset, and select and apply appropriate machine learning methods. You will structure and communicate the analysis using CRISP-DM. The project should critically reflect on assumptions, limitations, and opportunities for improvement, and—where relevant to the chosen case—consider issues such as deployment, organizational implications, ethics, governance, or risk.

Schedule

Project work begins with the first course session, continues through the presentation phase, and concludes with the final submission.

gantt
    dateFormat YYYY-MM-DD
    axisFormat %d %b
    todayMarker off

    section Proposal
    Form group and develop proposal   :proposal, 2026-09-01, 2026-09-10
    Proposal deadline                 :milestone, 2026-09-10, 0d
    Topic review                      :review, 2026-09-11, 7d

    section Analysis
    Case, data, modeling & evaluation :analysis, 2026-09-18, 2026-10-01
    Presentation submission           :milestone, 2026-10-01, 0d

    section Presentation
    Presentation period               :present, 2026-10-02, 2026-10-09
    Presentation 1                    :milestone, 2026-10-02, 0d
    Presentation 2                    :milestone, 2026-10-09, 0d

    section Finalization
    Revise and complete report        :revision, 2026-10-10, 2026-11-02
    Final submission                  :milestone, 2026-11-02, 0d

Group formation and topic proposal

  • Form your own group of 2–3 participants.

  • The group must send an email to g.wagner@fs.de by 2026-09-10 containing:

    • the names of all group members;
    • a concise proposed project topic; and
    • the concrete business or analytical question the group intends to address.

  • Proposed topics will be reviewed during the following week. The instructor may request adjustments if topics overlap too strongly or the proposed scope is unsuitable.

The following are illustrative examples, not predefined project options:

  • Finance, such as credit default, fraud, or customer lifetime value
  • Marketing, such as churn, customer segmentation, or campaign response
  • Human Resources, such as attrition, promotion, or team productivity
  • Digital Business, such as user engagement, gig worker retention, or demand prediction

Propose a concrete business problem and analytical question rather than merely selecting a generic domain.

Analysis task

  • Define a practical business problem or decision situation and formulate a concrete analytical question.
  • Create a synthetic but realistic dataset representing internal organizational data. You may complement it with synthetic or real external data where appropriate. Generate the synthetic data reproducibly in a separate R script.
  • Develop an integrated Quarto analysis document (.qmd) that presents the complete CRISP-DM analysis, including explanatory text, R code, executed results, and necessary figures or tables.
  • Present and critically reflect on the findings, their connection to course content, simplifying assumptions and limitations, and opportunities for improvement.
  • Complete one meaningful extension beyond a straightforward reproduction of classroom techniques. This may concern methods, data preparation, model comparison or robustness, deployment, organizational implications, ethics or governance, or communication. It does not need to involve a more complex machine-learning algorithm.
NoteNotes
  • Support key elements of your analysis with reputable sources, particularly the relevance of the problem, typical data sources, and common modeling approaches.
  • Simplifications are acceptable if they are clearly stated and justified. Explain how a more realistic analysis could extend beyond them.
  • If you plan to work with a large-scale big-data, real-time, or streaming-data scenario, consult with me in advance because of the additional technical complexity.
  • If questions come up during the project phase, you can arrange a meeting via Bookings and join meetings via Teams.

Presentation

Presentations take place on 2026-10-02 and 2026-10-09. Each group receives a 30-minute presentation slot followed by up to 15 minutes for questions and discussion. Every group member must present part of the project and contribute to the discussion; attendance is required for all project presentations.

All groups must submit their presentation PDF and relevant analysis files by 2026-10-01 20:00, irrespective of their assigned presentation slot. Submit them via email.

Submit the presentation as a PDF that:

  • clearly introduces the business problem and decision context;
  • explains the analytical approach and key modeling decisions;
  • communicates the most important findings and their practical relevance; and
  • critically discusses limitations, risks, the extension, and potential improvements.

Final submission

Submit the final project by 2026-11-02 via the Canvas assignment. The required files are specified below:

  • The R script used to generate the synthetic dataset together with the resulting dataset file(s), normally as CSV files. The script should document the business purpose, schema, generation logic, assumptions, distributions, relationships, and realism features such as noise or missing values where appropriate. Use a fixed random seed so that the dataset can be reproduced from the script, and keep the generation logic separate from the main .qmd.

  • The main report as a Quarto document (.qmd) together with its rendered PDF. The .qmd is the central analytical artifact and must contain the complete CRISP-DM analysis, explanatory text, R code, executed results, and the figures and tables needed to support the argument. Do not create a second, redundant report narrative. The main report text is limited to 12 pages. The title page, references, and appendices are excluded from this limit. Include enough code and results to make the work reproducible while keeping the prose focused. Submit both the .qmd and rendered PDF. The submitted source, data, and supporting files must reproduce the PDF.

TipIndividual contributions

Each student should be the primary author of approximately three pages of report text. The remaining material may be written jointly or developed collaboratively. Include a brief contribution statement at the end of the report identifying the sections or subsections primarily written by each group member. Every group member must also contribute meaningfully to the analysis, presentation, and discussion.

Evaluation

The final group evaluation is primarily based on the submitted project materials and is worth 120 performance points.

The presentation is a mandatory formative milestone rather than a separately graded component. Use the presentation and discussion to test your argument, identify weaknesses, and obtain feedback for improving the final report.

Participation in the presentation and discussion may additionally be considered when assessing individual contributions.

Category Points Criteria
A. Case and dataset design 30 pts Clarity and relevance of the problem; realism of the dataset; transparency of data generation
B. Analysis quality 40 pts Method justification; correctness; CRISP-DM use; insightfulness
C. Communication 40 pts Structure; visualization; clean code; argument quality; use of references; reflection
D. Extension 10 pts Meaningful differentiation beyond straightforward reproduction of classroom examples

AI policy

  • Allowed: debugging code and creating the R script for synthetic dataset generation
  • Not allowed: generating complete end-to-end solutions
  • You must be able to explain and defend your work at any time
  • All outputs must be validated

Example: Synthetic data generation in R

set.seed(42)
n <- 1000

# Generate related variables and add realistic noise.
age <- rnorm(n, mean = 40, sd = 10)
income <- age * 1000 + rnorm(n, mean = 0, sd = 5000)

# Add missing values where they are appropriate to the scenario.
income[runif(n) < 0.05] <- NA_real_

dataset <- data.frame(age, income)
write.csv(dataset, "dataset.csv", row.names = FALSE)

Adapt and document the generation logic for your own business context.

NoteNote

High-quality project work may contribute to future course development, for example by extending teaching materials or informing teaching cases. In such cases, we will discuss how contributions are acknowledged.