13. Privacy, Ethics, Fraud and Advertising Governance

Introduction: Performance Under Conditions of Trust

Digital advertising has traditionally been described as a system of targeting, delivery, measurement and optimisation. That description is incomplete. Advertising also involves the collection of personal information, the classification of individuals, the design of persuasive environments, the use of automated decisions and the distribution of commercial messages through complex supply chains. Each of these activities creates obligations. A campaign may be efficient in terms of clicks and conversions yet still be unlawful, discriminatory, deceptive, unsafe or socially harmful.

Privacy, ethics, fraud and governance are therefore not peripheral compliance topics. They are strategic conditions of sustainable performance marketing. Consumer trust affects willingness to share data, respond to advertising and remain loyal to a brand. Regulatory intervention can interrupt campaigns, impose financial and reputational costs, and restrict business models. Invalid traffic can inflate apparent performance, while poor viewability and unsafe placements can make media expenditure ineffective. Governance connects these issues by establishing who may collect data, who may make decisions, which suppliers may be used, how risks are monitored and what evidence is required before a campaign is approved.

This chapter examines responsible advertising as a managerial and operational discipline. It treats consent, data protection, tracking restrictions, fairness, manipulation, child protection, political advertising, accessibility, brand safety, viewability and ad fraud as interconnected issues rather than isolated checklists.

Learning Objectives

After studying this chapter, students should be able to:

1. Explain the relationship between privacy, ethics, performance and consumer trust.

2. Distinguish consent, transparency, legitimate business purposes and user control in data-driven advertising.

3. Evaluate tracking, targeting and measurement practices against principles of fairness, proportionality and accountability.

4. Identify dark patterns, manipulative techniques, discriminatory outcomes and risks affecting children and vulnerable audiences.

5. Analyse brand-safety, viewability and ad-fraud risks across the digital advertising supply chain.

6. Design a governance framework involving policies, roles, vendor controls, monitoring and incident response.

7. Apply responsible measurement practices when identifiers are restricted or data is incomplete.

Why Governance Has Become a Performance Issue

Performance marketing depends on feedback. An advertiser observes an impression, click, conversion or customer value and uses that information to improve future decisions. The feedback system is only useful when its inputs are lawful, sufficiently accurate and ethically obtained. A conversion attributed to an advertisement may be technically recorded but still be misleading if the user did not understand the tracking, if the impression was generated by a bot, or if the campaign excluded a protected group.

The central governance problem is an information asymmetry. Platforms, advertisers and intermediaries often know more about data collection and optimisation than the audience does. This asymmetry becomes problematic when it prevents meaningful choice or enables exploitation. A privacy notice hidden in legal language may formally disclose processing without creating genuine understanding. A consent interface that makes acceptance prominent and refusal difficult may obtain a click without respecting autonomy. Governance seeks to reduce this imbalance through transparency, limits, accountability and independent review.

A useful strategic test is to ask four questions about any advertising activity:

• Is it permitted by applicable law and platform policy?
• Is it understandable and reasonably expected by the audience?
• Is it fair in its effects on different groups?
• Is the measurement evidence reliable enough to support the decision?

A positive answer to one question cannot compensate for failure on the others.

A Vocabulary for Responsible Advertising

Privacy concerns the appropriate collection, use, disclosure and retention of information about identifiable or reasonably identifiable people. Personal data may include obvious identifiers such as an email address, but also device identifiers, location signals, browsing behaviour, inferred interests and combinations of apparently non-identifying data.

Consent is a person’s voluntary, specific, informed and unambiguous agreement to a defined processing activity. It is not equivalent to silence, pre-ticked boxes or acceptance of general terms. Consent must also be withdrawable, and withdrawal should not be made unnecessarily difficult.

Data minimisation means collecting and using only information reasonably necessary for a stated purpose. Purpose limitation means that data gathered for one purpose should not automatically be repurposed for unrelated advertising. Storage limitation requires retention to be restricted to a justified period. These principles are reflected in many data-protection regimes, although legal terminology and obligations vary by jurisdiction.

First-party data is collected directly by an organisation through its own customer relationship or digital properties. Second-party data is another organisation’s first-party data shared through an agreement. Third-party data is obtained from external sources, often through aggregators or intermediaries. The distinction does not by itself determine whether use is lawful or ethical; provenance, permission, accuracy and purpose remain important.

An identifier is a signal used to recognise a browser, device, account or person. Cookies, mobile advertising identifiers, login information, IP addresses, hashed emails and probabilistic signals are examples. The removal or restriction of one identifier does not eliminate privacy risk if other signals are combined to recreate recognition.

Profiling involves analysing personal data to evaluate or predict characteristics, interests or behaviour. Automated decision-making occurs when software makes or substantially influences a decision without meaningful human intervention. These practices require special care where they affect access to opportunities, prices, credit, employment, housing, education or other significant interests.

An audience segment is a group defined by selected characteristics or behaviours. A proxy variable is a feature that appears neutral but may reproduce a sensitive characteristic. For example, location, language, device type or browsing patterns may indirectly correlate with ethnicity, income or age.

Consent, Transparency and Responsible Data Use

A responsible data practice begins with a data map. The advertiser should identify what data is collected, from whom, through which technology, for what purpose, where it is stored, who receives it, how long it is retained and what rights individuals have. The map should include pixels, software development kits, analytics tags, customer-data platforms, conversion APIs, data clean rooms and agency or platform integrations.

Transparency should be layered. A short explanation at the point of collection can state what will happen and why; a fuller notice can provide details about recipients, retention and rights. The objective is not to eliminate legal documentation but to make the relevant decision intelligible at the moment it occurs. People should not have to infer from vague statements that their activity will be used for personalised advertising or shared with numerous vendors.

Consent management platforms can record choices and transmit signals to advertising partners. They do not, however, make an activity compliant merely by displaying a banner. The design, wording, default settings, vendor list, purpose categories and withdrawal process all matter. Where consent is not the appropriate legal basis, the organisation must still establish a lawful and fair rationale and respect applicable objection rights.

Privacy by design incorporates protection before a campaign is launched. Examples include contextual targeting instead of individual profiling, aggregated reporting instead of user-level logs, short retention periods, limited access privileges, encryption and suppression of sensitive audiences. A privacy impact assessment is useful where processing is extensive, involves vulnerable people, combines datasets or introduces new technology.

Data quality is also an ethical issue. Incorrect demographic or interest labels can lead to irrelevant advertising, exclusion or harmful assumptions. Organisations should provide mechanisms for correction, avoid treating inferred characteristics as facts and periodically test whether segments remain accurate and necessary.

Tracking Restrictions and Measurement Without Surveillance

Browsers, operating systems and regulators have increasingly limited cross-site tracking, third-party cookies, persistent identifiers and opaque fingerprinting. These changes challenge last-click attribution and individual-level retargeting, but they do not prevent rigorous measurement. They encourage advertisers to distinguish decision usefulness from unnecessary data collection.

First-party measurement can include authenticated interactions, consented analytics and server-side event transmission. Server-side methods may reduce exposure of raw data in the browser, but they do not remove obligations: the organisation remains responsible for the data it sends and the purposes for which it is used. Clean rooms can allow controlled analysis between parties without directly revealing individual records, yet they require strict governance against re-identification and excessive data enrichment.

Aggregated methods include media-mix modelling, geo-experiments, incrementality tests, conversion lift studies and controlled holdouts. These methods estimate whether advertising caused additional outcomes rather than simply assigning credit to a trackable touchpoint. Their limitations should be documented. A model may be directionally useful while remaining uncertain at campaign or audience level.

A sound measurement hierarchy is often preferable to a single attribution number:

1. Verify delivery and quality: was the ad served, viewable and placed in an acceptable environment?

2. Measure attention or interaction where relevant: did people meaningfully engage?

3. Estimate incremental response through experiments or robust modelling.

4. Connect outcomes to business value, while reporting uncertainty and assumptions.

This approach prevents precision from being confused with truth. A dashboard showing two decimal places does not make an estimate exact.

Ethics: Autonomy, Fairness and the Boundaries of Persuasion

Advertising is persuasive by nature, but persuasion becomes ethically questionable when it defeats informed choice or exploits vulnerability. Ethical analysis asks not only whether a message is legal, but whether its method is respectful, proportionate and defensible if disclosed publicly.

Autonomy requires that people can understand the commercial nature of a message and make decisions without coercion or concealed pressure. Scarcity claims, emotional appeals, personalised timing and social proof may be legitimate when truthful. They become manipulative when fabricated, impossible to decline, targeted at a known vulnerability or designed to bypass deliberation.

Dark patterns are interface designs that steer users toward an outcome they might not otherwise choose. Examples include preselected advertising consent, visually dominant acceptance buttons, confusing cancellation flows, disguised advertisements, repeated prompts after refusal and misleading countdown timers. Dark patterns are not simply poor user experience; they can undermine consent and produce unfair commercial outcomes.

Fairness has at least three dimensions. Procedural fairness concerns whether people are treated by a transparent and consistent process. Allocative fairness concerns the distribution of opportunities, prices or exposure. Representational fairness concerns whether groups are stereotyped, erased or portrayed with dignity. No single fairness metric resolves every conflict. A campaign may achieve equal delivery across groups while still using harmful creative portrayals, or achieve equal click rates while systematically excluding a group from an opportunity.

Advertisers should test both inputs and outcomes. Input review examines targeting criteria, exclusions, training data and optimisation goals. Outcome review examines reach, frequency, cost, creative exposure and conversion rates across relevant groups. Statistical differences are signals for investigation, not automatic proof of discrimination; equally, an absence of measured difference is not proof of fairness when data is incomplete.

Children, Vulnerable Audiences and Political Communication

Children generally have less ability to understand commercial intent, long-term data consequences and persuasive design. Advertising directed at children should therefore use age-appropriate explanations, avoid manipulative collection practices and comply with applicable rules concerning parental permission, profiling, behavioural advertising, sensitive data and targeted content. Age assurance is difficult: collecting excessive identity information to establish age can create a new privacy risk. A proportionate approach should consider the service, the likely audience, the risk and the least intrusive method available.

Vulnerable adults may include people experiencing financial distress, addiction, illness, bereavement or cognitive limitations. Sensitive inferences should not be used to intensify exploitation. For example, targeting people inferred to be financially desperate with high-cost credit or gambling promotions presents serious ethical and reputational risks even where a platform technically permits the audience.

Political advertising requires heightened scrutiny because it can influence civic participation, public trust and democratic debate. Governance should address sponsor identification, funding disclosure, targeting restrictions, record keeping, archival requirements, claims substantiation and separation of campaign communications from editorial content. Synthetic media and generative AI make provenance especially important. A political message that is lawful in one jurisdiction may be restricted in another, so campaign activation requires geographic legal review rather than a single global rule.

Accessibility and Inclusion as Advertising Quality

Accessibility is both an ethical obligation and a quality dimension. Advertising should be perceivable, operable, understandable and robust across devices and assistive technologies. Practical measures include captions for video, meaningful alternative text, sufficient colour contrast, readable typography, keyboard navigation, clear language, transcripts, non-flashing creative and accessible landing pages. Accessibility should be tested in the complete journey, not only in the advertisement: a captioned video that leads to an unusable form remains exclusionary.

Inclusive advertising also considers representation, language, cultural context and the possibility that automated optimisation will favour audiences easiest to convert. Reach and performance reports should reveal who is not being reached when exclusion affects access to products, information or opportunities. Inclusion is not achieved by adding diverse imagery while retaining targeting or product practices that disadvantage the same groups.

Brand Safety, Suitability and the Advertising Supply Chain

Brand safety concerns preventing advertising from appearing alongside content that creates unacceptable legal, reputational or social risk. Brand suitability is more contextual: content may be acceptable for one brand, audience or campaign and inappropriate for another. Risk categories can include terrorism, graphic violence, hate, misinformation, adult content, illegal activity, unsafe user-generated content and controversial news.

The programmatic supply chain may involve advertisers, agencies, demand-side platforms, exchanges, supply-side platforms, publishers, verification firms and resellers. Every additional intermediary creates uncertainty about inventory, data access, fees and accountability. Domain lists alone are insufficient because a safe domain can host unsafe content, and unsafe environments can evade simple keyword filters through coded language or images.

Controls should combine pre-bid rules, post-bid verification, inclusion and exclusion lists, contextual classification, publisher agreements, placement reporting and escalation procedures. The advertiser should define tolerance levels in advance rather than deciding after a crisis. Suitability controls should also account for adjacency, creative tone, frequency and the sensitivity of the product.

Viewability, Invalid Traffic and Ad Fraud

An impression is not necessarily an opportunity to see. Viewability measures whether an ad was rendered in a viewable portion of a screen for a minimum time under an accepted measurement standard. It remains a proxy rather than proof of attention: a viewable ad can be ignored, obscured or displayed to an automated browser.

Invalid traffic includes activity that does not represent genuine user interest. General invalid traffic may result from crawlers, data-centre traffic or accidental technical processes; sophisticated invalid traffic may imitate human behaviour, generate fake conversions or manipulate auctions. Ad fraud can involve domain spoofing, hidden ads, ad stacking, pixel stuffing, click farms, botnets, arbitrage, fake apps and falsified supply-path information.

Fraud creates more than wasted media expenditure. It contaminates models, causes teams to optimise toward artificial signals and can expose brands to unsafe or illegal environments. Detection requires a combination of independent verification, log-level analysis, anomaly detection, placement transparency, frequency analysis, conversion-quality checks and supplier accountability. No fraud tool detects everything; vendors should explain methodology, coverage, false-positive handling and audit rights.

A practical control sequence is to establish a baseline, identify unusual patterns, quarantine questionable inventory, investigate source paths, adjust buying rules and record the financial and measurement impact. Refunds should not be treated as the only remedy. The organisation must determine why controls failed and whether the same supplier should remain approved.

A Governance Architecture for Advertising Organisations

Governance converts principles into repeatable decisions. A mature framework usually contains five connected layers.

The first is policy. Policies define acceptable data practices, prohibited targeting, creative standards, children’s advertising rules, political advertising procedures, brand-safety categories, accessibility expectations and fraud tolerance. They should distinguish mandatory legal requirements from organisational risk appetite.

The second is accountability. A responsible organisation assigns roles to marketing, privacy or legal teams, information security, procurement, analytics, media operations and senior leadership. A responsibility matrix can specify who approves a data source, who evaluates a vendor, who monitors delivery and who can pause a campaign. Accountability should remain with the advertiser even when execution is outsourced.

The third is control design. Controls include consent records, access management, encryption, data retention schedules, pre-launch reviews, audience suppression, creative approvals, domain verification, brand-safety filters, fraud monitoring and incident response. Controls should be proportionate to risk and tested rather than assumed to work.

The fourth is evidence. Organisations should retain records of data provenance, legal basis, audience definitions, model assumptions, vendor contracts, approvals, test results, delivery reports, complaints and remediation. Documentation enables audit and prevents organisational memory from depending on individual employees.

The fifth is learning. Complaints, near misses, blocked impressions, rejected creatives, measurement discrepancies and incidents should produce improvements. A blame-oriented culture encourages concealment; a learning culture treats early reporting as a control strength while still assigning responsibility for negligence.

Vendor governance deserves particular emphasis. Due diligence should examine ownership, subcontractors, data flows, security controls, certifications where relevant, model training practices, deletion processes, audit rights, breach notification, indemnity, service levels and exit arrangements. Contracts should prohibit undisclosed data reuse, require compliance with campaign restrictions and permit suspension when risk thresholds are exceeded. A vendor approved for analytics is not automatically suitable for audience enrichment or generative creative production.

Artificial intelligence adds further governance needs. An advertising team should document the model’s purpose, training-data provenance where available, human review, known limitations, testing for bias, use of personal data, output ownership and procedures for correcting harmful or false content. Automated bidding should have budget limits, anomaly alerts and a human override. Generative systems should not be allowed to invent product claims, testimonials, political statements or sensitive audience inferences without review.

Illustrative Case: The “Growth at Any Cost” Campaign

Background

A fictional consumer-finance company, Northstar Credit, appoints an agency to acquire new customers through programmatic advertising. The campaign uses a customer list, look-alike modelling, retargeting and automated bidding. The agency reports a low cost per application and recommends increasing the budget.

Situation and Problem

After three months, complaints rise. Some people receive high-cost loan advertisements repeatedly after visiting websites about debt counselling. A consumer group alleges that the campaign disproportionately reaches financially distressed users. An audit also finds that a significant share of conversions came from a small set of mobile applications with low viewability and unusual click patterns. The consent records for one data partner are incomplete.

Analysis

The campaign has four separate but connected weaknesses. First, the data partner cannot demonstrate that its audience segments were collected for advertising purposes with an appropriate permission or lawful basis. Second, the optimisation system treats applications as successful conversions without examining approval quality, repayment risk or customer harm. Third, the retargeting strategy uses sensitive contextual signals as a proxy for financial vulnerability. Fourth, the low cost per application is inflated by invalid traffic and poor inventory quality.

The agency’s dashboard is not false in a narrow technical sense, but it is strategically misleading because it reports an incomplete outcome. The campaign has optimised a proxy rather than responsible business value.

Approach and Solution

Northstar pauses the questionable data source and retargeting segment, preserves relevant logs and begins an incident review. Legal and privacy teams assess notification and remediation obligations. Media operations introduce app-ads verification, independent invalid-traffic measurement, minimum viewability thresholds and a curated supply path. The marketing team replaces application volume with a quality-adjusted outcome that includes approved, suitable and sustainable customers.

The company adopts a prohibited-segment policy for inferred financial distress, requires documented provenance for partner data, and introduces a pre-launch review involving compliance, analytics and customer-experience teams. Creative is rewritten in plain language, with clear cost disclosures and no artificial urgency. Reporting now includes reach, frequency, viewability, invalid traffic, data-source status, complaints and uncertainty around attribution.

Outcome

The revised campaign produces fewer applications but a more reliable and sustainable customer portfolio. Reported efficiency initially worsens because invalid activity is removed. Over subsequent reporting periods, the company gains stronger evidence about incremental acquisition and reduces complaints. The case demonstrates that governance can change the definition of performance rather than merely add administrative burden.

Learning Points

1. A low acquisition cost is not sufficient evidence of success.

2. Sensitive outcomes can arise from apparently ordinary contextual and behavioural signals.

3. Vendor provenance and measurement quality must be assessed together.

4. Pausing a campaign can be an appropriate performance decision when evidence is unreliable.

5. Responsible optimisation requires business-quality and customer-impact measures, not only platform conversions.

Common Failure Modes and Managerial Choices

A frequent mistake is treating compliance as a one-time approval. Data flows, platform policies, vendors and creative variants change during a campaign, so governance requires continuous monitoring. Another mistake is assuming that anonymisation is absolute. Combining datasets, rare attributes or persistent identifiers can make re-identification possible.

Teams also confuse platform permission with ethical acceptability. A platform may allow an audience category that is inappropriate for a particular product or jurisdiction. Similarly, a privacy-safe identifier can still support manipulative targeting if the underlying purpose is harmful.

Overreliance on a single dashboard creates false confidence. Advertisers should compare platform reports with independent verification, server logs, experiment results and customer outcomes. They should investigate dramatic improvements rather than celebrating them automatically; fraud and tracking errors often appear as unusually efficient performance.

Another failure is using generic blacklists without understanding context. Broad exclusions may protect a brand but also suppress legitimate journalism, minority publishers or important public information. Suitability requires calibrated categories, human review for ambiguous cases and periodic evaluation of reach and representation.

Governance can also become excessively restrictive. If every uncertainty leads to zero experimentation, organisations may lose the ability to learn. The appropriate response is controlled experimentation: define the purpose, minimise data, set a duration and risk threshold, test outcomes, and stop when assumptions fail.

Current Developments and Responsible Practice

The advertising ecosystem is moving toward a mixed measurement environment: more first-party relationships, contextual signals, aggregated reporting, clean-room collaboration, consented authentication and experimental methods. None is a universal replacement for previous tracking. Each shifts the balance among accuracy, reach, cost, privacy and control.

Privacy-enhancing technologies may reduce direct exposure of user-level data through aggregation, local processing, secure computation and controlled matching. Their effectiveness depends on implementation and governance; technical architecture cannot compensate for an unjustified purpose or excessive retention.

Generative AI is accelerating creative production, translation, personalisation and synthetic imagery. It also raises risks of fabricated claims, copyright disputes, discriminatory representation, deepfakes, undisclosed persuasion and insecure use of customer data. Human review, provenance records, approval thresholds and clear labelling are essential in high-risk contexts.

Responsible advertising practice can be summarised as follows:

• Define the purpose and risk before selecting the technology.
• Collect the least data needed and retain it for the shortest justified period.
• Make choices understandable, reversible and free from coercive design.
• Test targeting and outcomes for exclusion, proxy discrimination and harmful effects.
• Use independent verification for viewability, fraud and supply quality.
• Make vendors contractually and operationally accountable.
• Report uncertainty, limitations and non-delivery alongside positive results.
• Establish pause, escalation and remediation procedures before launch.
• Treat accessibility and inclusion as requirements of campaign quality.
• Review governance when products, platforms, laws or audience conditions change.

Chapter Summary

Privacy, ethics, fraud and advertising governance are integrated dimensions of performance marketing. Responsible advertising begins with clear purpose, lawful and meaningful data practices, transparency and user control. Tracking restrictions encourage a shift from surveillance-oriented attribution toward first-party, aggregated and experimental measurement. Ethical practice requires attention to autonomy, fairness, manipulation, vulnerable groups, children, political communication and accessibility. Brand safety and suitability protect the relationship between media context and brand meaning, while viewability and fraud controls protect both expenditure and evidence. A governance architecture must assign accountability, control vendors, document decisions, monitor outcomes and support incident response. The strategic objective is not merely to avoid penalties but to produce trustworthy, inclusive and decision-useful performance.

Key Terms

Consent: Voluntary, informed, specific and unambiguous agreement to a defined processing activity.

Data minimisation: Limiting data collection and use to what is necessary for a stated purpose.

Purpose limitation: Restricting personal data use to purposes communicated and justified at collection.

First-party data: Information collected directly by an organisation through its own relationship or properties.

Profiling: Analysing data to evaluate or predict a person’s characteristics, interests or behaviour.

Automated decision-making: A decision made or substantially influenced by software without meaningful human intervention.

Dark pattern: An interface design that unfairly steers users toward a choice or outcome.

Proxy variable: An apparently neutral attribute that may reproduce a sensitive characteristic.

Fairness: The ethical and organisational requirement that processes, outcomes and representation do not unjustifiably disadvantage groups.

Privacy by design: Incorporating privacy safeguards into systems and processes before deployment.

Contextual targeting: Selecting advertising environments or content without relying primarily on individual profiles.

Clean room: A controlled technical environment for analysing data between parties while limiting direct data disclosure.

Brand safety: Protection against advertising appearing in environments that create unacceptable risk.

Brand suitability: The contextual assessment of whether an environment is appropriate for a particular brand or campaign.

Viewability: The extent to which an ad is rendered in a measurable, viewable screen area for a defined period.

Invalid traffic: Impressions, clicks or actions that do not represent genuine user interest.

Ad fraud: Deliberate manipulation of advertising delivery, measurement or payment for illicit gain.

Supply path: The sequence of intermediaries through which digital inventory is bought and sold.

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

Governance: The policies, roles, controls, evidence and oversight used to direct and account for organisational activity.

Review Questions

Short Answer Questions

1. Why is platform permission not sufficient evidence that a targeting practice is ethical?

2. Distinguish first-party data from third-party data.

3. What is the difference between brand safety and brand suitability?

4. Why can a viewable impression still fail to demonstrate attention?

5. State two purposes of a privacy impact assessment.

Long Answer Questions

1. Explain how meaningful consent should be designed and governed in a performance-marketing campaign.

2. Discuss the relationship between tracking restrictions and alternative measurement methods such as incrementality testing and media-mix modelling.

3. Analyse how dark patterns can undermine both consumer autonomy and data-protection objectives.

4. Describe a governance architecture for managing advertising vendors, including accountability, evidence and incident response.

5. Explain why ad fraud damages not only media expenditure but also optimisation and strategic decision-making.

Analytical / Application Questions

1. An advertiser observes that a campaign’s cost per acquisition falls sharply after adding a behavioural segment supplied by an external vendor. What evidence should the advertiser request before scaling the campaign?

2. A political advertiser wants to use location, browsing behaviour and look-alike modelling to persuade undecided voters. Develop a risk-assessment process covering transparency, targeting, claims, data use and documentation.

3. A children’s educational app proposes personalised ads funded by a third-party SDK. Evaluate the proposal and recommend a safer data, creative, measurement and governance approach.

22. Multiple Choice Questions with Answers

Question 1

Which practice best reflects data minimisation in digital advertising?

A. Collecting every available signal in case it becomes useful later

B. Collecting only the data necessary for a defined advertising purpose

C. Sharing all collected data with approved vendors

D. Retaining user-level data indefinitely for modelling

Correct Answer: B. Collecting only the data necessary for a defined advertising purpose

Question 2

Why does a consent-management platform not by itself guarantee compliant consent?

A. It cannot transmit any signals to advertising partners

B. It eliminates the need for privacy notices

C. The interface, purposes, defaults and withdrawal process also determine whether choice is meaningful

D. Consent is never relevant to digital advertising

Correct Answer: C. The interface, purposes, defaults and withdrawal process also determine whether choice is meaningful

Question 3

Which measurement approach most directly estimates whether advertising caused additional outcomes?

A. Last-click attribution

B. A controlled holdout or incrementality experiment

C. Counting impressions only

D. Increasing the attribution window

Correct Answer: B. A controlled holdout or incrementality experiment

Question 4

Which example is most clearly a dark pattern?

A. A clearly labelled option to reject non-essential cookies

B. A short explanation followed by a detailed privacy notice

C. A cancellation process that requires repeated confusing steps while subscription is easy to start

D. A captioned video advertisement

Correct Answer: C. A cancellation process that requires repeated confusing steps while subscription is easy to start

Question 5

Why can location or device type create fairness risks even when sensitive characteristics are not directly collected?

A. They may act as proxy variables for sensitive characteristics

B. They always produce identical advertising outcomes

C. They prevent all forms of measurement

D. They are never personal data

Correct Answer: A. They may act as proxy variables for sensitive characteristics

Question 6

Which control is most relevant to detecting sophisticated ad fraud?

A. Using only a broad domain blacklist

B. Comparing log-level activity and conversion quality with independent verification

C. Increasing the campaign frequency cap without investigation

D. Removing all performance metrics

Correct Answer: B. Comparing log-level activity and conversion quality with independent verification

Question 7

What is the principal distinction between brand safety and brand suitability?

A. Safety concerns legal and reputational risk, while suitability is contextual to the brand and campaign

B. Safety concerns clicks, while suitability concerns impressions

C. Safety applies only to television advertising

D. Suitability eliminates the need for content classification

Correct Answer: A. Safety concerns legal and reputational risk, while suitability is contextual to the brand and campaign

Question 8

Which response best demonstrates responsible use of an automated bidding system?

A. Allowing unrestricted optimisation because the algorithm is objective

B. Removing human review after initial launch

C. Using budget limits, anomaly alerts, documented objectives and a human override

D. Optimising only for the cheapest recorded conversion

Correct Answer: C. Using budget limits, anomaly alerts, documented objectives and a human override

Question 9

Why is a low cost per application potentially misleading in the Northstar Credit case?

A. Applications are never useful business outcomes

B. The metric ignored customer quality, possible harmful targeting and invalid traffic

C. Programmatic advertising cannot generate applications

D. Viewability always increases acquisition cost

Correct Answer: B. The metric ignored customer quality, possible harmful targeting and invalid traffic

Question 10

Which practice best supports accessibility in a digital advertising journey?

A. Using text embedded in images without alternatives

B. Providing captions, sufficient contrast, keyboard access and an accessible landing page

C. Targeting only users with the newest devices

D. Replacing all written information with animation

Correct Answer: B. Providing captions, sufficient contrast, keyboard access and an accessible landing page