Assignment 4: Prompt engineering and a marketing use-case
Systematic prompt experimentation, then a critical reflection: where would text as data actually create value in a marketing scenario — and what could go wrong?
The task
The closing assignment is part experiment, part judgement — no heavy compute.
Part 1 — Prompt engineering. Using an LLM platform of your choice, start from “Write a product description for a new type of running shoe” and systematically vary the prompt: add a target audience, then create at least five variations across length (short vs. detailed), tone (formal, conversational, persuasive), and perspective (“as a sales representative”, “as a sports journalist”). Record how each change moves the output, and analyse the patterns: what do specificity, tone, and context each buy? How could prompt engineering support customer engagement and brand messaging?
Part 2 — Text as data for marketing strategy. Choose a concrete marketing scenario where text as data adds value (customer sentiment, personalised content, chatbots, trend analysis, …) and:
- Describe the use case and its business context
- Compare two course tools applied to it — benefits and weaknesses of each
- Pick one and outline an implementation plan: tools, process, data sources, success metrics
- Reflect critically: ethics, bias, privacy, and the limits of interpretation — what could go wrong, and how would you know?
Materials
- Notebook: assignment4.ipynb — the full task text (this is a written assignment; there is no code to run)
- Any LLM interface works for part 1 — note which model and date you used, since outputs vary
Deliverable
A written report covering both parts, with your prompt variations and outputs documented in part 1. Groups of 1–3; part of the collected end-of-semester submission. Format details on Itslearning.
Builds on
Lecture 11 — the prompting checklist in its backup slides is a good companion — and the whole course’s toolbox for part 2.