AI, Digital Transformation, and the Future of Media

Introduction

Advertising and media planning are being reshaped by artificial intelligence, platformisation, connected devices, privacy regulation, and new forms of data collaboration. Media planning was traditionally organised around relatively stable channels—television, radio, print, outdoor, and desktop display—and relied on panel research, demographic segments, negotiated prices, and periodic campaign reports. Contemporary planning is more dynamic. Audiences move across devices and platforms; inventory is bought in milliseconds; creative assets can be generated and adapted automatically; and campaign decisions are increasingly informed by predictive models.

This transformation does not mean that human media planners are becoming unnecessary. Rather, their work is moving upward in value: from manually placing advertisements toward defining objectives, evaluating evidence, governing data and algorithms, designing experiments, managing risk, and making strategic judgements. The central professional challenge is to combine technological capability with public trust, brand suitability, legal compliance and human accountability.

This chapter examines the principal technologies and practices shaping the future of media: generative AI, predictive targeting, automated creative, algorithmic buying, synthetic media, clean rooms, first-party data, privacy-enhancing technologies, connected television, retail media, immersive media, attention metrics, and responsible human oversight.

Learning Objectives

By the end of this chapter, students should be able to:

1. Explain how AI and digital transformation are changing media planning and buying.
2. Distinguish generative AI, predictive targeting, automated creative and algorithmic media buying.
3. Evaluate first-party data, clean rooms and privacy-enhancing technologies.
4. Assess opportunities and limitations in connected television, retail media and immersive media.
5. Use reach, frequency, response and attention measures appropriately.
6. Identify risks involving privacy, bias, misinformation, brand safety, intellectual property and accountability.
7. Design a media-planning approach that combines automation with responsible human oversight.

1. From Digitisation to Intelligent Media Systems

Digitisation converts analogue content or processes into digital form. Digital transformation is broader: it changes the organisation’s operating model, customer relationships, decision processes and sources of value. In media, transformation includes addressable distribution, platform-based buying, real-time measurement, automated optimisation, cloud data infrastructure and AI-assisted content production.

An intelligent media system can be understood as a connected arrangement of:

• Data: audience signals, contextual information, transaction records, content metadata and outcome data.
• Models: statistical or machine-learning systems that estimate likely outcomes.
• Decision rules: constraints concerning budget, frequency, brand suitability, geography and privacy.
• Activation systems: demand-side platforms, retail-media platforms, connected-TV systems, social platforms and creative-delivery tools.
• Measurement and governance: experiments, reporting, audits, human review and accountability mechanisms.

The quality of such a system depends on more than model accuracy. A highly accurate model trained on unlawful or biased data may create unacceptable outcomes. Similarly, an efficient buying system may optimise cheap impressions rather than valuable attention or incremental sales. Media management therefore requires a multi-dimensional definition of performance: effectiveness, efficiency, reach, quality, safety, fairness, compliance and long-term brand value.

2. Generative AI and the Changing Creative Process

Generative AI produces text, images, audio, video or code in response to prompts or structured inputs. Large language models generate text by estimating probable sequences of tokens; image and video systems learn patterns from training data and generate new outputs from prompts, reference images or other conditions. In advertising, applications include:

• drafting headlines, scripts and product descriptions;
• creating multiple language or format variations;
• producing storyboards and mood boards;
• resizing and adapting assets to platform specifications;
• generating synthetic voice-overs or background imagery;
• assisting research synthesis and campaign reporting.

Generative AI is best treated as a production and ideation assistant rather than an autonomous source of truth. It can produce fluent but inaccurate statements, reproduce stereotypes, reveal confidential information, or generate material that resembles protected creative work. A responsible workflow includes a defined brief, approved source material, prompt and output records, human fact-checking, accessibility review, legal review and final sign-off.

The creative process is consequently becoming more modular. Instead of producing one advertisement for a mass audience, a team may create a library of approved claims, images, calls to action, product feeds and brand rules. Software then assembles variants for particular placements or contexts. This can increase relevance and testing speed, but excessive variation may fragment brand identity and make causal evaluation more difficult.

Automated Creative Optimisation

Automated creative optimisation selects or assembles creative variants according to predicted performance. A system may compare headlines, images, formats, offers or calls to action. It should distinguish between:

• Personalisation: adapting content to a known or inferred audience characteristic.
• Contextual adaptation: adapting content to the surrounding content, time, location or device without identifying an individual.
• Dynamic creative optimisation: automatically assembling approved components.
• Generative production: creating new material rather than selecting from a fixed library.

A practical control is to maintain a “creative constitution”: a documented set of non-negotiable rules covering claims, tone, visual identity, prohibited representations, accessibility, cultural sensitivity, disclosures and escalation procedures. Automation should operate inside this constitution, not replace it.

3. Predictive Targeting and Algorithmic Buying

Predictive targeting uses historical and contextual data to estimate the probability of a future outcome, such as a site visit, completed view, purchase or subscription. The system does not necessarily know that a person belongs to a fixed demographic category; it may infer that a device, household or context resembles cases associated with a desired outcome.

Important distinctions include:

• Descriptive analytics: what happened?
• Predictive analytics: what is likely to happen?
• Prescriptive optimisation: what action should be taken under stated constraints?

Algorithmic buying uses software to select and purchase advertising inventory. In programmatic auctions, a buyer may submit a bid when an impression becomes available. The decision can consider predicted value, price, placement, frequency, audience or context. The speed of automation creates scale, but also introduces opaque supply chains, invalid traffic, auction manipulation, data leakage and brand-safety risks.

A planner should ask five questions before trusting a targeting model:

1. What exactly is the prediction target?
2. What data generated the prediction, and was its use lawful and fair?
3. Does the model perform consistently across relevant groups and contexts?
4. What action follows from the prediction, and who bears the risk of error?
5. Can the result be tested against a credible counterfactual?

Measurement Foundations

Basic media measures remain useful when their definitions and denominators are clear.

Here, CTR is click-through rate, C is the number of recorded clicks and I is the number of served or measured impressions. Both C and I must refer to the same campaign scope and measurement rules. CTR measures recorded response, not necessarily attention, quality, incremental sales or profit.

ROAS is return on advertising spend, R is attributable revenue and M is advertising spend. Both must use consistent currency and attribution rules. ROAS is not the same as profit because it excludes, unless explicitly incorporated, product costs, fulfilment, overhead and the effect of other marketing activity.

For a controlled test, incremental lift can be expressed as:

Here, Y_T is the outcome rate or mean outcome for the treatment group and Y_C is the corresponding outcome for a comparable control group. The interpretation assumes that assignment and measurement are sufficiently sound, groups are comparable, and other influences do not differentially affect the groups. If Y_C is zero, percentage lift is undefined and an absolute difference should be reported instead.

For example, suppose a holdout test records a purchase rate of 3.0% among exposed users and 2.5% among the control group. The relative lift is (3.0 − 2.5) ÷ 2.5 × 100 = 20%. The absolute difference is 0.5 percentage points. Both figures should be reported because “20% lift” can sound larger than the underlying change.

4. Identity, First-Party Data and Privacy-Enhancing Technologies

First-party data is information collected directly by an organisation through its customer relationships and owned touchpoints, such as authenticated accounts, subscriptions, transactions, customer-service interactions, loyalty programmes and consented website activity. It can support relevance and measurement, but ownership of a database does not automatically create permission for every use. Purpose limitation, transparency, data minimisation, retention controls and withdrawal mechanisms remain important.

Second-party data is another organisation’s first-party data shared through an agreement. Third-party data is generally collected or aggregated by an entity that does not have a direct relationship with the individual. Legal definitions differ by jurisdiction, and contractual labels do not override applicable privacy law.

A clean room is a controlled computing environment in which organisations compare or analyse datasets without freely exchanging raw identifiable records. For example, an advertiser and a media platform may compute overlap, reach or conversion outcomes inside a clean room and receive aggregated results subject to minimum reporting thresholds. Clean rooms may reduce unnecessary data movement, but they are not automatically anonymous or risk-free. Governance must address access, query design, re-identification risk, retention, permitted outputs and independent audit.

Privacy-enhancing technologies (PETs) are technical methods intended to reduce exposure of personal information while allowing useful analysis. Examples include:

• Aggregation: reporting groups rather than individuals.
• Pseudonymisation: replacing direct identifiers with tokens; this does not make data anonymous if re-identification remains possible.
• Differential privacy: adding calibrated statistical noise so that the presence or absence of one person has limited influence on a published result.
• Federated learning: training models across decentralised data stores while sharing model updates rather than centralising raw data.
• Secure multi-party computation: enabling parties to compute a result jointly without revealing their private inputs.
• Trusted execution environments: processing data in hardware-protected areas with controlled access.

PETs involve trade-offs among privacy, accuracy, cost, latency and explainability. They should complement, not replace, legal review, ethical assessment and data governance.

5. Connected Television and the Reorganisation of Video

Connected television (CTV) refers to television content delivered through an internet-connected television, streaming device, game console or similar system. Connected-TV advertising may be bought through direct publisher agreements, platform marketplaces or programmatic channels. It combines television-like sight, sound and motion with digital delivery and, in some settings, more flexible audience and outcome measurement.

CTV creates opportunities for:

• household-level or contextual addressability, subject to permissions and platform rules;
• frequency coordination across streaming services;
• sequential storytelling;
• more rapid reporting than traditional broadcast measurement;
• integration with web, mobile or retail outcomes.

Its limitations include fragmented reach, inconsistent definitions of a completed view, co-viewing uncertainty, identity duplication, limited cross-platform transparency, subscription environments with restricted advertising and concerns about frequency management. A CTV plan should not assume that one device equals one person or that a logged impression equals human attention.

6. Retail Media and Commerce-Connected Planning

Retail media is advertising sold by retailers or commerce platforms using their digital properties, physical environments or transaction-linked audiences. Formats include sponsored search, product listings, onsite display, offsite audience extension, in-store screens and digital circulars. Its strategic attraction is proximity to purchase intent and access to sales data.

Retail media can shorten the path between exposure and transaction, but it presents important questions. Retailers may become both marketplace operators and media sellers, creating possible conflicts of interest. Measurement may favour platform-attributed sales, which can overstate incremental impact if shoppers were already likely to buy. Advertisers should request clear definitions of exposure, attribution windows, organic cannibalisation, new-to-brand customers and incrementality.

A robust retail-media programme separates three outcomes:

1. Media delivery: whether the advertisement was served in an appropriate environment.
2. Behavioural response: whether searches, visits, add-to-baskets or conversions changed.
3. Business incrementality: whether the campaign caused additional profitable sales beyond what would otherwise have occurred.

7. Immersive Media and Synthetic Media

Immersive media includes augmented reality (AR), virtual reality (VR), mixed reality, spatial computing and interactive three-dimensional environments. Advertising applications include virtual product trials, location-based overlays, interactive demonstrations, branded virtual spaces and experiential commerce.

Immersive campaigns can create richer involvement, but novelty is not a strategy. The experience must provide user value, be accessible, function on the intended devices and respect bodily, spatial and behavioural data. Motion, gaze, voice and biometric-like signals may be especially sensitive. Clear consent and restraint are essential.

Synthetic media is media generated or materially altered by AI, including synthetic presenters, cloned voices, generated people and manipulated video. It can reduce production costs and enable localisation, but it can also facilitate impersonation, deceptive endorsements, non-consensual likeness use and misinformation. Disclosure should be considered when the synthetic nature could affect audience interpretation, trust or safety. Consent, rights clearance, provenance records and secure asset management are fundamental.

8. Attention Metrics and the Limits of Exposure

An impression indicates delivery according to a measurement system; it does not guarantee that a person noticed, understood or remembered an advertisement. Attention metrics attempt to measure or model indicators such as viewable time, active time, gaze, interaction, completion, sound status, screen prominence or survey-based recall.

Attention can be measured directly in controlled studies, estimated through behavioural signals, or modelled using media and creative characteristics. Each approach has different validity, cost and scalability. A gaze measure may indicate visual attention but not persuasion. A long dwell time may reflect confusion rather than engagement. Therefore, attention should be connected to outcomes through validation rather than treated as a universal currency.

A useful measurement hierarchy is:

• Delivery: was the opportunity to see technically available?
• Quality: was the placement viewable, audible where relevant, fraud-reduced and brand-suitable?
• Attention: did evidence suggest active processing or engagement?
• Effect: did awareness, preference, behaviour or sales change?
• Value: did the change justify cost and support long-term objectives?

Planners should avoid replacing one simplistic metric with another. A balanced dashboard might include reach, effective frequency, viewability, completed views, attention indicators, brand lift, incremental conversions, cost per incremental outcome, complaints and carbon or production measures.

9. Responsible Human Oversight

Responsible oversight is not a final ceremonial approval. It is a system of controls distributed across planning, data acquisition, modelling, activation, creative production, measurement and post-campaign review.

Core Governance Principles

Lawfulness and consent: Data collection and activation must comply with applicable privacy, consumer-protection, copyright, advertising and AI rules. Consent must be meaningful where required, and refusal should not be engineered through manipulative interface design.

Purpose limitation: Data collected for service delivery should not silently become an advertising dataset. New uses require a documented compatibility and legal assessment.

Fairness: Targeting and exclusion can reproduce social inequalities. Teams should test whether protected or vulnerable groups are disproportionately denied opportunities, exposed to harmful content or charged different offers.

Explainability and contestability: People affected by significant decisions should receive understandable information and a route for correction or appeal where applicable.

Security: Access controls, encryption, retention limits, incident response and vendor due diligence are necessary because advertising systems contain commercially and personally sensitive information.

Provenance and intellectual property: Teams should record where data and creative assets came from, what licences apply, and whether generated content uses a person’s likeness or voice with permission.

Brand and social safety: Keyword lists alone are inadequate. Contextual classification, inclusion and exclusion rules, supply-path controls, human escalation and continuous monitoring are required.

Human accountability: An organisation cannot transfer responsibility to a platform or model. A named decision-maker should own objectives, constraints, approval and remediation.

A Practical Human-in-the-Loop Model

A useful model has four stages:

1. Human framing: define the business problem, public-interest constraints, eligible data, prohibited uses and success criteria.
2. Machine assistance: allow systems to generate, predict, rank or optimise within documented boundaries.
3. Human challenge: inspect assumptions, subgroup performance, creative claims, placements, costs and unexpected outputs.
4. Human decision and learning: approve, pause or reject activity; document reasons; conduct experiments and update controls.

Human oversight can fail if reviewers are overloaded, lack expertise or merely approve automated recommendations. Effective oversight requires authority, time, training, audit trails and incentives that reward responsible interruption rather than only delivery volume.

10. An Integrated Future Media Planning Framework

An advanced media plan can follow this sequence:

Step 1: Define the objective and causal question

State whether the objective is awareness, consideration, trial, sales, retention or another outcome. Ask what must change and what evidence would demonstrate that the campaign caused or contributed to the change.

Step 2: Map the audience without over-collecting data

Combine broad strategic understanding with contextual signals, consented first-party relationships and carefully governed partner data. Do not assume that more personal data creates better planning.

Step 3: Select channels for their role

Assign CTV, retail media, search, social, audio, outdoor, immersive and other channels according to reach, context, attention, credibility, actionability and cost—not merely fashionable technology.

Step 4: Build an approved creative system

Create modular assets, claims, product feeds, accessibility specifications and brand rules. Identify which variations may be automated and which require human review.

Step 5: Activate with transparent controls

Set budget limits, frequency rules, placement exclusions, supply-path requirements, consent signals, model constraints and escalation thresholds. Require vendors to explain fees, data use and measurement methodology.

Step 6: Measure delivery, quality and incrementality

Use platform reporting for operational monitoring, independent verification where appropriate, and experiments or credible quasi-experimental designs for causal questions. Avoid treating attributed conversions as automatically incremental.

Step 7: Review societal and organisational effects

Assess privacy incidents, bias indicators, complaints, accessibility, misinformation risk, production rights, environmental impact and the effect on professional roles.

Applied Case: Illustrative Scenario—Aurora Foods’ Responsible Commerce Campaign

Background

Aurora Foods, a fictional national packaged-food brand, wants to increase trial of a new low-sugar breakfast product. It has a loyalty programme, a retailer partnership and a library of professionally produced video and display assets. The marketing team proposes generative copy, predictive targeting, CTV, retail media and automated creative optimisation.

Problem

The initial proposal optimises for attributed online sales. It uses a broad data match with a retail partner, creates many AI-generated variants and buys inventory across several exchanges. The privacy officer is concerned about unclear permissions; the brand team is concerned about unsupported health claims; and the finance team questions whether reported sales are incremental.

Analysis

The team reframes the campaign around three outcomes: qualified reach, product trial and incremental household purchases. It identifies the following risks:

• a purchase predictor may favour existing loyal buyers rather than new trialists;
• retail-platform attribution may claim sales that would have happened anyway;
• generated copy may imply medical benefits;
• excessive variants may weaken brand consistency;
• cross-platform frequency may be miscounted;
• partner data may be used beyond the purpose communicated to customers.

Approach and Solution

Aurora uses consented first-party loyalty data for suppression and measurement, while contextual and broad audience approaches provide prospecting scale. The retailer and Aurora conduct an aggregated clean-room analysis with minimum cell sizes and no export of individual records. A control or holdout design is agreed before activation.

Generative AI drafts alternative headlines, but only pre-approved nutritional claims can enter the production library. Human reviewers verify every claim, image, voice and disclosure. A fixed set of creative combinations is tested rather than allowing unlimited generation.

CTV is used for reach and video storytelling; retail media supports high-intent search and product-page visibility; contextual display and audio extend reach. Frequency rules are set within each channel, while the team reports that cross-platform frequency is only partially observable.

The campaign dashboard distinguishes served impressions, quality-adjusted delivery, completed views, attention indicators, product-page visits, trial purchases and incremental purchases. A brand-lift study assesses awareness and consideration, while the holdout estimates sales incrementality.

Outcome

The campaign produces fewer attributed sales than the original platform forecast but demonstrates a positive incremental effect among prospecting households. The fixed creative system generates enough variation for testing without weakening brand consistency. The clean-room process improves retailer collaboration while limiting raw-data movement. Some CTV reach remains difficult to deduplicate, so the team reports uncertainty rather than presenting a falsely precise total.

Learning Points

1. The best optimisation target is not always the easiest metric to count.
2. First-party data is valuable only when its collection and use are properly governed.
3. Generative AI increases production capacity but does not remove claim verification.
4. Clean rooms reduce certain risks but require strong controls and careful interpretation.
5. Incrementality, not platform attribution alone, should guide investment decisions.
6. Explicit uncertainty is a professional strength, not a reporting weakness.

Challenges and Future Directions

Several developments will continue to test media managers. AI agents may negotiate inventory, generate plans and execute optimisation with limited human prompting. Regulation may impose stronger duties concerning automated decision-making, transparency, provenance and high-risk applications. Identity signals may become more fragmented, increasing the importance of contextual intelligence and consented relationships. Retail and media companies may converge, raising competition and governance questions. Spatial computing may create new forms of advertising based on location, gaze and environment. At the same time, economic pressure may encourage automation faster than organisations can build oversight.

Future-ready professionals therefore need hybrid competence: media economics, audience research, statistics, experimentation, data governance, creative judgement, vendor management, privacy literacy and ethical reasoning. The strategic question is not whether a technology is new, but whether it creates defensible value for audiences, advertisers and media owners without imposing unacceptable costs on society.

Best-Practice Checklist

Before launching an AI-enabled media campaign, confirm that:

• the objective and causal question are explicit;
• the data source, permission and retention period are documented;
• targeting does not unlawfully or unfairly exploit sensitive characteristics;
• generated and personalised content is fact-checked and rights-cleared;
• brand-safety, accessibility and misinformation controls are active;
• vendors disclose relevant fees, data practices and measurement limitations;
• frequency and identity limitations are reported honestly;
• a control, experiment or other credible evaluation method is planned;
• human owners have authority to pause the campaign;
• incidents, decisions and model changes are recorded for audit.

Chapter Summary

AI and digital transformation are changing media planning from periodic channel selection into a continuously measured, algorithmically assisted system. Generative AI supports ideation and production, while predictive targeting estimates likely outcomes and automated creative assembles approved variations. Algorithmic buying expands scale and speed but introduces opacity, fraud, data and brand-safety risks.

First-party data can strengthen customer relationships, but permission and purpose remain essential. Clean rooms and privacy-enhancing technologies can enable collaboration while reducing raw-data movement, although they do not eliminate re-identification or governance risks. Connected television combines video impact with digital delivery; retail media links advertising to commerce data; immersive and synthetic media create new experiences while raising consent, safety, provenance and intellectual-property concerns.

Exposure, attention, response and incremental business value are different constructs. Sound planning reports them separately and uses experiments where causal claims matter. The future of media requires human-in-the-loop governance: people must frame objectives, challenge model outputs, approve consequential decisions and remain accountable for outcomes. Technological sophistication is valuable only when matched by transparency, fairness, security, lawful data use and respect for audiences.

Key Terms

Generative AI: AI that creates new text, images, audio, video or other content from learned patterns and instructions.

Predictive targeting: Using data and models to estimate which audiences or contexts are likely to produce a desired outcome.

Automated creative optimisation: Software selection or assembly of creative variants according to defined performance objectives.

Algorithmic buying: Automated selection and purchase of advertising inventory using rules, data and models.

Synthetic media: Media generated or materially altered by artificial intelligence.

First-party data: Data collected directly by an organisation through its own customer or audience relationships.

Clean room: A controlled environment for analysing datasets while restricting access to raw individual-level information.

Privacy-enhancing technology: A technical method that reduces personal-data exposure while supporting legitimate computation or analysis.

Connected television: Television content delivered through internet-connected devices or streaming systems.

Retail media: Advertising sold through or by a retailer using its digital, physical or commerce-related assets.

Immersive media: AR, VR, mixed-reality or spatial experiences that increase interaction with a mediated environment.

Attention metric: A measure or estimate of active viewing, processing, interaction or engagement beyond basic delivery.

Incrementality: The additional outcome caused by advertising compared with what would have occurred without it.

Holdout group: A control group deliberately not exposed to a campaign for comparative measurement.

Brand safety: Practices that reduce the risk of advertisements appearing beside unsuitable or harmful content.

Data minimisation: Collecting and retaining only the data necessary for a specified purpose.

Differential privacy: A privacy method that limits the effect of any one person’s data on a published result, often through calibrated noise.

Human-in-the-loop: A system in which people retain defined review, intervention and accountability responsibilities.

Attribution: Assigning reported credit for an outcome to one or more marketing contacts or channels.

Provenance: Documented information about the origin, transformation and rights status of data or media.

Review Questions

Short Answer Questions

1. Distinguish digitisation from digital transformation in media.
2. What is the difference between predictive targeting and automated creative optimisation?
3. Why is first-party data not automatically lawful for every advertising purpose?
4. State two limitations of connected-TV measurement.
5. Why should attributed sales not automatically be interpreted as incremental sales?

Long Answer Questions

1. Explain how generative AI may change advertising production, and evaluate the controls required for responsible use.
2. Compare clean rooms with privacy-enhancing technologies. Discuss what risks they reduce and what risks remain.
3. Analyse the opportunities and limitations of retail media for advertisers, retailers and consumers.
4. Explain the hierarchy from delivery to business value and show why attention cannot replace outcome measurement.
5. Design a human-in-the-loop governance model for an AI-enabled media-planning department.

Analytical / Application Questions

1. A campaign records 48,000 clicks from 2,400,000 measured impressions. Calculate CTR and explain what the result does not prove.
2. In a holdout study, the exposed group has a purchase rate of 4.2% and the control group has a rate of 3.5%. Calculate relative lift and absolute percentage-point difference. State one assumption needed for a causal interpretation.
3. Using the Aurora Foods scenario, recommend how the team should combine CTV, retail media and contextual prospecting while controlling privacy, frequency, attribution and brand-safety risks.

Multiple Choice Questions with Answers

Question 1

Which statement best describes generative AI in advertising?

A. It only reports historical campaign performance

B. It creates or transforms content from learned patterns and instructions

C. It guarantees that every advertising claim is legally accurate

D. It replaces the need for creative strategy

Correct Answer: B. It creates or transforms content from learned patterns and instructions

Question 2

A campaign records 30,000 clicks from 1,000,000 measured impressions. What is its CTR?

A. 0.3%

B. 3%

C. 30%

D. 300%

Correct Answer: B. 3%

Question 3

What is the principal purpose of a clean room?

A. To publish every customer record to all campaign partners

B. To eliminate the need for privacy law

C. To analyse datasets in a controlled environment without freely exchanging raw records

D. To guarantee that all advertising outcomes are incremental

Correct Answer: C. To analyse datasets in a controlled environment without freely exchanging raw records

Question 4

Which is an example of a privacy-enhancing technology?

A. Secure multi-party computation

B. Unlimited retention of identifiable profiles

C. Unrestricted third-party cookie sharing

D. Manual duplication of customer databases

Correct Answer: A. Secure multi-party computation

Question 5

Why can retail-media attribution overstate campaign effectiveness?

A. Retail media never records transactions

B. The platform may receive credit for purchases that would have happened without the advertising

C. Retail media cannot target product searches

D. Retail media only measures brand awareness

Correct Answer: B. The platform may receive credit for purchases that would have happened without the advertising

Question 6

Which statement about connected television is most accurate?

A. One connected-TV device always represents one individual viewer

B. CTV combines internet delivery with video advertising opportunities but may have deduplication and co-viewing limitations

C. CTV cannot support audience or contextual decision-making

D. CTV measurement is identical across all platforms

Correct Answer: B. CTV combines internet delivery with video advertising opportunities but may have deduplication and co-viewing limitations

Question 7

What does incrementality measure?

A. The total number of impressions served

B. The additional outcome caused by advertising compared with a suitable counterfactual

C. The number of creative variants produced

D. The average length of an advertisement

Correct Answer: B. The additional outcome caused by advertising compared with a suitable counterfactual

Question 8

Which practice best supports responsible generative-AI creative production?

A. Publishing outputs without review to preserve speed

B. Allowing the model to invent health claims for testing

C. Using approved claims, rights checks, fact-checking and human sign-off

D. Removing all records of prompts and revisions

Correct Answer: C. Using approved claims, rights checks, fact-checking and human sign-off

Question 9

Why should attention metrics be interpreted cautiously?

A. Attention is unrelated to advertising exposure

B. A signal such as dwell time may indicate processing but does not by itself prove persuasion or sales impact

C. Attention can only be measured with sales data

D. Attention metrics always measure individual identity

Correct Answer: B. A signal such as dwell time may indicate processing but does not by itself prove persuasion or sales impact

Question 10

Which responsibility belongs within a human-in-the-loop system?

A. Allowing a platform to own all accountability for campaign outcomes

B. Defining objectives and constraints, challenging model outputs and retaining authority to pause activity

C. Ignoring subgroup performance when aggregate results are positive

D. Using automation without audit trails

Correct Answer: B. Defining objectives and constraints, challenging model outputs and retaining authority to pause activity