Paid Social, Display and Programmatic Media

Introduction

Digital advertising has moved from manually purchased space toward automated, data-informed systems that select audiences, creative, placements and prices in milliseconds. Paid social advertising uses platforms such as Meta, TikTok, LinkedIn and Pinterest to deliver sponsored messages within social environments. Display advertising uses visual formats across websites, applications and connected devices. Programmatic media uses software, audience data, automated auctions and algorithmic decision-making to buy many forms of digital inventory.

For brand managers, the central task is not simply to purchase impressions. It is to connect business objectives, audience strategy, creative design, media economics, measurement and governance. Effective campaigns reach relevant people at an appropriate level of frequency, generate useful outcomes and protect brand trust.

Learning objectives

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

1. Distinguish paid social, display and programmatic media.
2. Design audience segments and targeting strategies.
3. Select suitable creative formats and placements.
4. Explain auctions, bidding and media-buying models.
5. Plan retargeting while managing privacy and frequency.
6. Calculate and interpret core media metrics.
7. Evaluate attribution and optimisation choices.
8. Critically assess automation, AI, brand safety and regulation.

1. The paid-media system

Paid media is purchased exposure. It differs from owned media, which a brand controls, and earned media, which results from third-party attention or advocacy. Paid media can accelerate reach, create demand, support launches and return measurable performance data. However, it can also produce waste, consumer irritation, privacy risk and misleading performance claims.

The main participants are:

• Advertiser or brand: defines objectives, budget, audience and acceptable outcomes.
• Agency: may provide strategy, planning, buying, creative and measurement.
• Publisher or platform: supplies audience access and advertising inventory.
• Ad exchange: facilitates transactions between buyers and sellers.
• Demand-side platform (DSP): allows advertisers to buy inventory and apply targeting and bidding rules.
• Supply-side platform (SSP): helps publishers sell available inventory.
• Data providers: may supply contextual, demographic or behavioural signals, subject to law and consent.
• Measurement and verification providers: assess delivery, viewability, invalid traffic, brand safety and outcomes.

A useful planning sequence is: business objective → communication objective → audience → proposition and creative → inventory and buying method → measurement → optimisation. Reversing this sequence—for example, choosing a platform because it is fashionable—often produces activity without strategic value.

2. Paid social and display compared

Paid social ads appear in social feeds, stories, short-video environments, search results within social platforms, messaging spaces or creator placements. They benefit from platform interaction data, native creative formats and social-context signals. Their limitations include platform dependence, opaque algorithms, changing policies and possible overreliance on engagement metrics.

Display advertising traditionally refers to visual advertising on websites and applications. Common forms include banners, responsive ads, rich media, native placements, interstitials and video. Display can support broad reach and repeated exposure, while contextual placement can associate a brand with relevant content. Low attention, banner blindness, fraud and accidental clicks are important limitations.

Programmatic is a method of buying and delivering media, not a single format. Programmatic buying may include display, online video, audio, connected television and digital out-of-home inventory. It can use real-time bidding, private marketplaces, guaranteed contracts or automated direct deals.

3. Objectives, audiences and targeting

Campaign objectives should be expressed in business terms. Awareness may seek incremental reach or brand recall; consideration may seek qualified visits or product exploration; conversion may seek sales, registrations or applications; retention may seek repeat purchase or reactivation. The buying system should optimise toward an event that is both measurable and commercially meaningful.

Audience architecture

A practical audience plan distinguishes:

• Prospecting audiences: people who have not meaningfully interacted with the brand.
• Remarketing audiences: people who visited, viewed content, used an app or began but did not complete an action.
• Customer audiences: existing purchasers or account holders, used for retention, upselling or exclusion.
• Lookalike or modeled audiences: platform-created groups statistically similar to a source audience. Similarity is an estimate, not proof of shared needs.
• Contextual audiences: people reached because of the content or environment in which an advertisement appears, rather than because an individual profile has been inferred.

Segmentation should be actionable. A segment is useful when it has a distinct need, can be reached lawfully, can receive a relevant message and can be evaluated separately. Overly narrow targeting can raise costs, reduce delivery and amplify algorithmic bias. Broad targeting may be effective when a platform has strong conversion signals, but it requires sufficient budget, reliable measurement and appropriate creative variation.

Targeting signals can include geography, time, device, language, content context, search behaviour, declared interests, first-party interactions and purchase events. Sensitive characteristics and inferred personal information require special care and may be prohibited by platform rules or law. Exclusions are as important as inclusions: recent purchasers may be excluded from acquisition campaigns, employees from a recruitment test, and converted users from an uncompleted-action sequence.

4. Creative formats and message design

Creative and media are interdependent. A technically accurate audience cannot rescue a weak or irrelevant message. Creative decisions should reflect attention conditions, screen size, sound availability, brand codes and the desired action.

Common formats include:

• Single-image ads: efficient and clear, but dependent on strong visual hierarchy.
• Video: useful for demonstration, emotion and narrative; the opening seconds must communicate value.
• Carousel: supports multiple products, stages or benefits.
• Collection and catalogue ads: connect discovery with product browsing.
• Lead forms: reduce friction but may generate lower-intent submissions.
• Native advertising: adopts the surrounding editorial form and therefore requires clear disclosure.
• Rich media: permits interaction, but higher production complexity does not guarantee effectiveness.
• Responsive creative: allows a platform to assemble assets across placements; governance is needed to prevent unsuitable combinations.

Effective variants change one meaningful element at a time where learning is the goal: proposition, image, opening frame, call to action, audience or landing page. A campaign should preserve recognisable brand assets while adapting execution to each placement. Accessibility requires readable contrast, captions for video, meaningful alternative text where supported, and avoidance of flashing or misleading interaction cues.

5. Media buying and the auction

Digital inventory may be bought by negotiated price, fixed delivery, cost per thousand impressions (CPM), cost per click (CPC), cost per acquisition (CPA), cost per completed view or value-based bidding. The metric used for payment is not necessarily the metric that determines business success.

In a simplified impression-based purchase, CPM is:

Cost is the media expenditure in currency and impressions are delivered ad exposures. The factor 1,000 expresses the cost per thousand impressions. The formula assumes impressions are counted consistently; it does not establish that impressions were viewable, unique or non-fraudulent.

For example, a campaign costing £12,000 and delivering 2,400,000 impressions has:

CPM = (£12,000 ÷ 2,400,000) × 1,000 = £5.

CPC and CPA are similarly calculated:

These metrics are meaningful only when clicks and conversions are defined consistently and attribution rules are disclosed.

In real-time bidding, an impression opportunity is sent to an auction. Eligible buyers evaluate the opportunity, submit bids and apply pacing, budget, audience, placement and safety rules. The platform determines a winner according to its auction design and quality or relevance factors. The advertiser’s bid is not always identical to the final price. Auction mechanics vary, so media planners should rely on platform documentation and independent verification rather than assume that every market is a simple second-price auction.

A campaign budget can be allocated by daily or lifetime limits, with pacing systems attempting to distribute delivery over time. A high bid may increase access but also raise cost; a low bid may reduce delivery. The appropriate bid depends on expected value, conversion probability, marginal cost and strategic constraints.

6. Programmatic architecture and supply quality

A typical programmatic path is: user or device requests content → publisher makes an impression available → SSP or exchange sends an opportunity → DSP evaluates advertiser rules → auction selects an eligible bid → creative is checked and served → exposure and outcome signals are recorded. This process may occur in milliseconds.

Programmatic arrangements include:

• Open auction: broad access to inventory, usually with less control.
• Private marketplace: selected buyers receive access to specified inventory.
• Preferred deal: a buyer receives an opportunity at an agreed price before broader auction access.
• Programmatic guaranteed: inventory and price are negotiated while delivery is automated.

Supply-path optimisation seeks efficient and transparent routes between buyer and publisher. Important controls include allowlists and blocklists, seller identification standards, ads.txt or equivalent publisher declarations, app-ads.txt, domain verification, frequency rules, viewability requirements and independent fraud measurement. Cheap inventory can be expensive if it generates invalid traffic, poor attention or brand damage.

7. Frequency, retargeting and sequencing

Reach is the number of distinct people exposed to a campaign. Impressions are total exposures. Average frequency is:

For 900,000 impressions delivered to 300,000 people, average frequency is 3. This average hides distribution: one person may see an ad once while another sees it ten times. Frequency distributions and reach-by-frequency reports are therefore more informative than an average alone.

Excessive repetition can cause irritation, creative fatigue and inefficient spending. Frequency management can use caps, time windows, creative rotation, exclusion after conversion and sequential storytelling. A cap should be interpreted carefully because identity resolution differs across browsers, devices, logged-in environments and platforms.

Retargeting can be valuable because prior interaction indicates some level of interest. Typical sequences include awareness exposure, product education, reminder, proof or offer, followed by suppression after purchase. Retargeting should not exploit sensitive circumstances, reveal private behaviour or continue indefinitely. Consent, retention limits and clear privacy notices are essential. Contextual alternatives and first-party relationship strategies become increasingly important as third-party identifiers decline.

8. Measurement and attribution

A measurement plan should specify the unit of analysis, event definitions, reporting window, data source, quality controls and decision rule. Core metrics include reach, frequency, impressions, viewability, clicks, click-through rate (CTR), conversion rate, CPA, return on advertising spend (ROAS) and incremental outcomes.

CTR is:

If 18,000 clicks arise from 900,000 impressions, CTR = 2%.

ROAS is:

A ROAS of 4 means £4 of attributed revenue per £1 of advertising cost. It is not the same as profit: product cost, fulfilment, salaries and overhead may be excluded. A campaign with lower ROAS can be strategically preferable if it creates more incremental customers or stronger long-term value.

Attribution assigns credit for an outcome to one or more touchpoints. Last-click attribution gives all credit to the final measured interaction; first-click gives it to the initial interaction; linear attribution distributes credit equally; position-based models assign greater weight to selected positions; data-driven models estimate contribution using observed patterns. Each model is an assumption, not a direct observation of causality.

Problems include cookie loss, cross-device fragmentation, walled gardens, self-reporting, view-through claims, duplicate conversions and selection bias. A person who was likely to buy may be more likely to see or click an ad, creating an inflated impression of effectiveness. Incrementality tests—such as randomised holdouts, geo experiments or platform conversion-lift studies—ask what happened because of advertising compared with a credible counterfactual.

9. Optimisation and decision-making

Optimisation should follow a learning agenda rather than constant adjustment. First verify implementation: pixels or tags, event deduplication, consent signals, product feeds, URLs, conversion definitions and brand-safety settings. Then examine delivery, quality, response and commercial outcomes.

A useful diagnostic tree is:

1. Insufficient delivery: check bid, budget, audience size, approval, inventory and pacing.
2. Delivery without attention: check viewability, placement, load time and creative relevance.
3. Attention without action: check proposition, call to action, landing-page speed and message match.
4. Action without value: check lead quality, fraud, product economics and attribution.
5. Short-term value without growth: test incrementality, new audiences and longer-term brand outcomes.

Change one or a small number of variables in a controlled test. Avoid declaring a winner from tiny samples or repeatedly stopping and restarting learning systems. Statistical significance is not a substitute for managerial significance, and platform-reported results should be reconciled with independent analytics and business records.

Worked planning example

An online retailer spends £20,000. It receives 1,000,000 impressions, 15,000 clicks, 600 reported purchases and £72,000 attributed revenue.

• CPM = (£20,000 ÷ 1,000,000) × 1,000 = £20.
• CTR = (15,000 ÷ 1,000,000) × 100% = 1.5%.
• CPC = £20,000 ÷ 15,000 = £1.33 approximately.
• CPA = £20,000 ÷ 600 = £33.33 approximately.
• ROAS = £72,000 ÷ £20,000 = 3.6.

The results appear attractive, but a manager must still ask whether purchases were incremental, whether revenue is gross or net, whether returns were deducted, and whether the £33.33 CPA is below the firm’s allowable acquisition cost.

10. Governance, ethics and regulation

Advertising systems process or infer information about people, making governance a strategic responsibility. Requirements vary by jurisdiction, but common principles include lawful processing, transparency, purpose limitation, data minimisation, security, retention control and respect for access or deletion rights. Consent may be required for particular cookies, pixels or personalised advertising activities; it must not be assumed merely because a platform can technically target a user.

Important governance practices include:

• documenting purposes, data sources and retention periods;
• obtaining valid consent where required and respecting refusal signals;
• avoiding targeting based on protected or highly sensitive attributes;
• labelling sponsored content and influencer relationships clearly;
• providing truthful claims and accessible terms;
• reviewing automated decisions for bias and exclusion;
• controlling children’s advertising and age-sensitive categories;
• monitoring brand adjacency, misinformation, extremist content and unsafe user-generated material;
• using contracts and audits for agencies, platforms and data processors.

AI can generate creative variants, predict response, optimise bids and identify anomalous traffic. It can also reproduce biased training data, create deceptive synthetic content, make unexplainable exclusions or produce legally problematic claims. Human review, approved data sources, version control, disclosure and outcome auditing are necessary. Automation should increase disciplined learning, not eliminate accountability.

11. Illustrative case: Northstar Outdoor Equipment

Background

Northstar, a fictional mid-sized outdoor-equipment retailer, launches a waterproof commuter backpack. Its objective is to acquire profitable new customers while building awareness among urban professionals.

Problem

An initial campaign uses broad display inventory and retargets every site visitor for 90 days. The dashboard reports a 5.2 ROAS, but sales records show little increase in total orders. Frequency is high, placement quality is inconsistent, and many attributed purchases were already likely to occur.

Analysis

The team separates prospecting from remarketing, defines a 30-day purchase window, removes existing purchasers from acquisition, and introduces an incrementality test. It audits supply paths, blocks unsuitable placements and uses viewability and invalid-traffic verification. Creative is adapted into short captioned video for social, product demonstrations for video inventory and contextual placements around commuting and weather content.

Approach and solution

The budget is divided between prospecting, controlled remarketing and a small brand-lift test. Prospecting uses broad, privacy-compliant contextual and platform audiences, while remarketing uses a short sequence: product view, practical proof, then a reminder. Purchasers are suppressed. Frequency caps and creative rotation are applied. The team evaluates incremental orders, contribution margin and new-customer rate rather than ROAS alone.

Outcome

In this illustrative scenario, reported ROAS falls to 3.8 because inflated view-through credit is removed, but incremental orders rise and the proportion of new customers improves. The company keeps the campaign because incremental contribution is stronger, even though the headline platform metric is lower.

Learning points

The case demonstrates that more attribution is not necessarily more effectiveness. Audience separation, supply quality, frequency control, creative relevance and experimental measurement produce better managerial decisions than a single dashboard number.

12. Contemporary developments and best practice

The decline of third-party identifiers is encouraging investment in first-party data, contextual targeting, clean rooms, modeled measurement and consent-based relationships. Connected television and retail media extend programmatic principles into premium video and commerce environments. These developments create opportunities but also fragmented reporting and new concentration of power among large platforms and retailers.

A professional campaign should:

1. Start with a measurable business and communication objective.
2. Define audiences by need and actionability, not merely demographic labels.
3. Match format, message and placement to attention conditions.
4. Use a planned mix of prospecting, retention and exclusions.
5. Set frequency and fatigue rules before launch.
6. Verify delivery, viewability, fraud and brand safety independently where possible.
7. Use consistent conversion definitions and reconcile platform data.
8. Combine attribution with incrementality and longer-term brand measures.
9. Test systematically and avoid premature optimisation.
10. Document consent, data provenance, contracts and human accountability.

Chapter Summary

Paid social, display and programmatic media are connected but distinct forms of digital advertising. Paid social provides native platform environments; display provides visual inventory across digital properties; programmatic automates the buying of many inventory types. Effective planning connects business objectives to audience design, creative, auctions, supply quality and measurement.

Audience strategies should distinguish prospecting, remarketing, customer and contextual approaches while respecting privacy and avoiding excessive narrowness. Media metrics such as CPM, CPC, CTR, CPA and ROAS describe delivery or attributed outcomes, but they do not by themselves prove causality or profitability. Frequency management, retargeting controls and creative sequencing reduce waste and fatigue. Attribution models are useful interpretive tools, whereas incrementality testing provides stronger evidence of causal impact. Finally, automation and AI require governance, transparency, quality assurance and human responsibility.

Key Terms

Paid media: Purchased exposure used to deliver marketing messages.

Paid social: Sponsored advertising delivered within social-platform environments.

Display advertising: Visual advertising delivered on websites, applications and related digital properties.

Programmatic buying: Software-enabled buying and delivery of digital advertising inventory.

DSP: Demand-side platform used by advertisers to evaluate and buy inventory.

SSP: Supply-side platform used by publishers to sell inventory.

Ad exchange: Technology marketplace that facilitates transactions between buyers and sellers.

Prospecting: Reaching people who have not meaningfully interacted with the brand.

Remarketing: Advertising to people who previously interacted with a brand or property.

Contextual targeting: Selecting placements according to surrounding content or environment.

Frequency: Number of exposures delivered to an individual or average exposures across reached people.

Reach: Number of distinct people exposed to an advertisement or campaign.

CPM: Cost per thousand impressions.

CTR: Percentage of impressions that produce clicks.

CPA: Cost per attributed conversion or acquisition.

ROAS: Attributed revenue divided by advertising cost.

Attribution: Allocation of credit for an outcome among marketing touchpoints.

Incrementality: The additional outcome caused by advertising compared with a credible counterfactual.

Viewability: Whether an impression met a defined opportunity-to-see standard.

Brand safety: Controls intended to prevent advertising from appearing beside unsuitable content.

Review Questions

Short Answer Questions

1. What is the difference between paid social and display advertising?
2. Define a demand-side platform and state its role.
3. Why should customer exclusions be used in an acquisition campaign?
4. What does average frequency measure?
5. Why is ROAS not equivalent to profit?

Long Answer Questions

1. Explain how a brand manager should move from a business objective to a paid-media plan.
2. Compare open auctions, private marketplaces and programmatic guaranteed buying.
3. Discuss the benefits and risks of retargeting, including privacy and fatigue.
4. Explain why attribution models cannot automatically establish advertising causality.
5. Evaluate the role of AI in media buying and identify appropriate governance controls.

Analytical / Application Questions

1. A campaign spends £8,000 and delivers 400,000 impressions, 8,000 clicks and 200 purchases. Calculate CPM, CTR, CPC and CPA, then identify two limitations of interpreting these figures.
2. A retailer reports high platform ROAS but no increase in total sales. Design an incrementality test and state which variables should be controlled.
3. A video campaign has strong reach but declining completion rates and rising complaints. Diagnose possible causes and recommend changes to frequency, creative, audience and measurement.

Multiple Choice Questions with Answers

Question 1

Which statement best describes programmatic media?

A. A method limited exclusively to social-feed advertising

B. A software-enabled method of buying and delivering digital inventory

C. An unpaid form of customer advocacy

D. A fixed creative format used only for banners

Correct Answer: B. A software-enabled method of buying and delivering digital inventory

Question 2

A campaign costs £6,000 and delivers 1,500,000 impressions. What is its CPM?

A. £0.25

B. £2.50

C. £4.00

D. £40.00

Correct Answer: C. £4.00

Question 3

Which audience is most clearly a prospecting audience?

A. People who purchased yesterday

B. People who visited the checkout but did not buy

C. People who have not meaningfully interacted with the brand

D. Employees excluded from targeting

Correct Answer: C. People who have not meaningfully interacted with the brand

Question 4

What is the main purpose of a frequency cap?

A. To guarantee every user clicks an advertisement

B. To limit repeated exposure to an individual or identifiable user

C. To increase the number of conversion events recorded

D. To replace creative testing

Correct Answer: B. To limit repeated exposure to an individual or identifiable user

Question 5

Which metric is calculated as clicks divided by impressions?

A. CPA

B. ROAS

C. CTR

D. CPM

Correct Answer: C. CTR

Question 6

Why can last-click attribution overstate the value of a retargeting campaign?

A. It ignores every recorded conversion

B. It assigns credit to the final interaction even when the user was already likely to convert

C. It measures only impressions and never clicks

D. It always excludes paid search

Correct Answer: B. It assigns credit to the final interaction even when the user was already likely to convert

Question 7

Which approach provides the strongest direct evidence of incremental advertising effect?

A. Comparing two creative colours without a control group

B. Using the platform’s highest reported ROAS

C. A randomised holdout or credible controlled experiment

D. Counting total impressions

Correct Answer: C. A randomised holdout or credible controlled experiment

Question 8

What is contextual targeting based primarily on?

A. The content or environment in which an advertisement appears

B. A user’s verified purchase intention in every case

C. The advertiser’s profit margin

D. The number of employees in an agency

Correct Answer: A. The content or environment in which an advertisement appears

Question 9

A campaign generates £45,000 attributed revenue from £15,000 advertising cost. What is ROAS?

A. 0.33

B. 1.5

C. 3

D. 30

Correct Answer: C. 3

Question 10

Which practice is most appropriate for responsible AI-supported media buying?

A. Allowing automated targeting to use any available sensitive data

B. Removing all human review after launch

C. Auditing outputs for bias, quality, legality and performance

D. Treating modeled predictions as proven causal effects

Correct Answer: C. Auditing outputs for bias, quality, legality and performance