Chapter 24 Exercises: AI for Marketing and Customer Experience

Chapter 24 Exercises: AI for Marketing and Customer Experience

Section A: Recall and Comprehension

Exercise 24.1 Describe the three eras of marketing evolution presented in this chapter (intuition-driven, data-driven, AI-augmented). For each era, identify one defining characteristic and one key limitation.

Exercise 24.2 Define the five levels of the personalization maturity model. For each level, provide one real-world example not mentioned in the chapter.

Exercise 24.3 Explain the five layers of conversational AI architecture (NLU, dialog management, knowledge retrieval, response generation, escalation logic). Why is escalation logic arguably the most important design decision?

Exercise 24.4 Compare and contrast the six attribution models presented in the chapter (first-touch, last-touch, linear, time-decay, position-based, data-driven). Under what circumstances would you recommend each?

Exercise 24.5 What is the "creepy line"? Identify and explain the five factors that determine whether personalization feels helpful or invasive.

Exercise 24.6 Define customer lifetime value (CLV) and explain the difference between traditional formula-based CLV models and machine learning-based CLV models. What additional data inputs do ML models leverage?

Exercise 24.7 Explain the concept of incrementality testing. Why is it superior to simple before/after comparisons for measuring marketing AI ROI?

Section B: Application

Exercise 24.8: Personalization Maturity Assessment Select a brand you interact with regularly (as a customer). Assess their personalization maturity against the five-level model: - (a) At which level does the brand currently operate? Provide at least three pieces of specific evidence. - (b) What would the brand need to do — in terms of data infrastructure, technology, and organizational capability — to advance to the next level? - (c) Should the brand advance to the next level? Consider the cost, customer expectations, and competitive dynamics. Justify your recommendation.

Exercise 24.9: Chatbot Design Challenge You are designing an AI-powered chatbot for a premium hotel chain. The hotel's brand is built on personalized, high-touch service, and the leadership is concerned that a chatbot will feel impersonal. - (a) Design the chatbot's scope — which customer service tasks should the chatbot handle, and which should be reserved for human agents? Justify each allocation. - (b) Describe the escalation logic. What signals should trigger handoff to a human agent? - (c) How would you handle a guest who is angry about a billing error? Write the first three exchanges of the conversation (chatbot and guest), demonstrating emotional awareness and appropriate escalation. - (d) How would you measure the chatbot's success? Identify at least five metrics, covering both efficiency and customer satisfaction.

Exercise 24.10: Attribution Analysis A direct-to-consumer skincare brand runs the following marketing mix: - Instagram ads (awareness) - Google search ads (consideration) - Email marketing (nurture) - Influencer partnerships (social proof) - Retargeting display ads (conversion)

A typical customer journey involves 4-6 touchpoints across these channels before purchasing. The CMO currently uses last-touch attribution and allocates 60 percent of the marketing budget to retargeting display ads, which "drive" the most conversions under this model.

(a) Explain why last-touch attribution likely overvalues retargeting and undervalues Instagram and influencer marketing.
(b) How would a data-driven attribution model likely redistribute credit? What data would you need to build such a model?
(c) Design an incrementality test to determine the true causal impact of retargeting ads. Include: treatment group, control group, duration, key metrics, and potential confounding factors.
(d) What organizational resistance might you encounter when presenting results that reduce retargeting's attributed credit? How would you manage this?

Exercise 24.11: Dynamic Pricing Ethics An online grocery delivery service uses AI-powered dynamic pricing. Consider each scenario and evaluate whether the pricing practice is ethical, unethical, or ambiguous. Justify each answer. - (a) Prices for bottled water increase by 40 percent during a heat wave, reflecting increased demand. - (b) A customer who has been browsing the platform for 30 minutes without purchasing sees prices increase by 5 percent on all items in their cart, based on the model's assessment of purchase intent. - (c) Loyal customers who order weekly receive lower prices than new customers, as a retention incentive. - (d) Customers in affluent zip codes see prices 8 percent higher than customers in lower-income zip codes, based on willingness-to-pay models. - (e) Prices on perishable items decrease automatically as they approach expiration, reducing food waste. - (f) A customer who compares prices on a competitor's site before returning sees a lower price than a customer who did not comparison-shop.

Exercise 24.12: NK's Opt-In Tiers NK designed a three-tier opt-in personalization system for AthenaPlus (Basic, Personalized, Proactive). - (a) For each tier, describe in detail what data is used, what the customer experience looks like, and what value the customer receives in exchange for the data they share. - (b) NK was surprised that 41 percent of pilot members opted into Tier 3 (Proactive). Propose three hypotheses that might explain this higher-than-expected adoption rate. For each hypothesis, describe how you would test it. - (c) Design a Tier 4 that goes beyond NK's current system. What additional data would it use? What additional value would it provide? Would this tier cross the creepy line? Why or why not? - (d) A privacy advocate argues that any system that uses browsing data for personalization — even with opt-in consent — is manipulative because consumers cannot fully understand the implications of their consent. How would you respond?

Exercise 24.13: Social Listening Crisis Response Your company (a mid-size airline) deploys an AI-powered social listening system. At 2:47 PM on a Tuesday, the system detects the following anomaly: mention volume has spiked 800 percent in the last 15 minutes, negative sentiment is at 94 percent, and a video showing a passenger being denied boarding despite having a valid ticket is going viral on multiple platforms. - (a) Design the first 60 minutes of your crisis response. Who is notified? In what order? What information do they need? - (b) Draft the company's first public response (to be posted within 30 minutes of detection). Balance speed, empathy, and the constraint that you do not yet know all the facts. - (c) What AI tools from this chapter and previous chapters could assist in managing this crisis? Identify at least three, explaining how each would be used. - (d) After the crisis is resolved, what systemic changes would you implement to prevent recurrence? Consider both operational and AI-system changes.

Section C: Analysis and Evaluation

Exercise 24.14: The Personalization Paradox Research consistently shows that (1) customers expect personalized experiences and reward brands that deliver them, while (2) customers are increasingly concerned about data privacy and distrust brands that collect too much data. This is the personalization paradox. - (a) Using examples from the chapter, explain how this paradox manifests in practice. - (b) Evaluate NK's three-tier opt-in model as a solution to the paradox. What are its strengths and limitations? - (c) Propose an alternative approach to resolving the paradox that does not rely on opt-in tiers. Consider privacy-preserving technologies (federated learning, differential privacy), contextual approaches, or other strategies. - (d) Is the personalization paradox fundamentally resolvable, or is it an inherent tension that must be managed rather than solved? Argue one side.

Exercise 24.15: The Manipulation Question At the end of the chapter, Ravi asks NK: "Where's the line between helpful personalization and manipulation?" - (a) Define manipulation in the context of marketing AI. How does it differ from persuasion? - (b) Identify three specific techniques described in this chapter that could be used manipulatively. For each, explain the line between legitimate persuasion and manipulation. - (c) Should companies be required to disclose when AI is making marketing decisions that affect individual customers? What would such a disclosure look like in practice? - (d) Design an "ethical marketing AI" framework — a set of principles and practices that a company could adopt to ensure its marketing AI remains persuasive without becoming manipulative. Include at least five principles with implementation guidelines.

Exercise 24.16: First-Party Data Strategy The chapter describes the collapse of third-party cookies as "the most significant structural shift in digital advertising since the invention of programmatic buying." - (a) Explain why first-party data is more valuable than third-party data in the post-cookie era. Consider quality, consent, exclusivity, and regulatory risk. - (b) Design a first-party data strategy for a mid-size fashion e-commerce company. What data collection mechanisms would you implement? What value exchange would you offer customers in return for their data? - (c) How does Athena's loyalty program function as a first-party data asset? What additional data collection opportunities does the program create? - (d) Critics argue that first-party data strategies simply move surveillance from third-party platforms to brands themselves, without fundamentally improving consumer privacy. Evaluate this argument.

Exercise 24.17: CLV and Bias The chapter warns that CLV models trained on historical data may "systematically undervalue customers from demographics that have historically received less marketing attention." - (a) Explain the mechanism by which this bias operates. Use a specific example. - (b) How would you test whether a CLV model exhibits this type of bias? Describe a methodology. - (c) If you discover that your CLV model assigns systematically lower predicted values to customers in a particular demographic group, what should you do? Consider both the business and ethical implications. - (d) Connect this issue to the concepts in Chapter 7 (supervised learning classification) and Chapter 9 (unsupervised learning clustering). How do training data biases in those upstream models flow through to NK's personalization engine?

Section D: Integration and Synthesis

Exercise 24.18: Full-Stack Marketing AI Design You are the newly appointed VP of Marketing at a direct-to-consumer furniture company with $200 million in annual revenue, 1.5 million customers, and no AI capabilities. Design a two-year marketing AI roadmap. - (a) Prioritize five AI marketing use cases from this chapter. Rank them by impact, feasibility, and strategic importance. Justify your ranking. - (b) For your top-priority use case, detail: data requirements, technology stack, team composition, estimated timeline, expected ROI, and key risks. - (c) How would you measure success? Design a metrics dashboard with at least eight metrics across revenue, engagement, satisfaction, and ethical health categories. - (d) What governance structures would you put in place to ensure your marketing AI is deployed responsibly? Reference concepts from this chapter and look ahead to Part 5.

Exercise 24.19: Competitive Analysis — Sephora vs. Cambridge Analytica Using the two case studies from this chapter: - (a) Create a comparison matrix evaluating both organizations across: data collection practices, transparency, customer value exchange, personalization approach, ethical boundaries, and outcomes. - (b) Identify three specific design decisions that Sephora made correctly that Cambridge Analytica made incorrectly (or vice versa). For each, explain the decision, its consequences, and the lesson for marketing AI practitioners. - (c) Is the distinction between Sephora and Cambridge Analytica primarily about technology, ethics, regulation, or business model? Argue your position. - (d) Apply the lessons from both case studies to NK's AthenaPlus project. Identify two ways NK's design reflects Sephora's positive example and one potential vulnerability where, if poorly managed, it could drift toward Cambridge Analytica's mistakes.

Exercise 24.20: The Part 4 Integration This chapter is the final chapter of Part 4. Reflect on how the concepts from Chapters 19-24 connect: - (a) Identify three specific ways that prompt engineering (Ch. 19-20) enables the marketing AI applications described in this chapter. - (b) How does the RAG architecture from Chapter 21 apply to marketing chatbots? - (c) What role do no-code/low-code platforms (Ch. 22) play in democratizing marketing AI? - (d) How does the cloud AI services evaluation framework from Chapter 23 apply to selecting a marketing AI technology stack? - (e) Looking ahead to Part 5, identify three ethical and governance questions raised by this chapter that the remaining chapters will need to address.

Selected answers are provided in Appendix B: Answers to Selected Exercises.