Paid Search and Campaign Architecture

Introduction

Paid search is the practice of purchasing qualified visibility on search-engine results pages and related networks. Unlike organic search engine optimization, which seeks unpaid prominence through relevance, authority and technical quality, paid search uses advertising platforms, bids, budgets, targeting settings and creative assets to participate in real-time auctions. The principal platforms include Google Ads, Microsoft Advertising and specialist marketplace-search systems such as Amazon Ads. Although interfaces and policies differ, their underlying managerial problem is similar: connect a user’s commercial intent with an appropriate message and landing experience while controlling cost and risk.

Campaign architecture is the deliberate arrangement of an advertising account into portfolios, campaigns, ad groups, keywords, ads, audiences and conversion actions. Good architecture makes performance measurable and controllable. Poor architecture creates mixed intent, weak relevance, unreliable reporting and inefficient budgets. Architecture is therefore not merely an administrative task; it is a form of strategic segmentation.

Learning Objectives

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

1. Explain the principal paid-search platforms, formats and auction mechanisms.
2. Design an account, campaign and ad-group structure aligned with business objectives and user intent.
3. Select keyword match types and construct effective negative-keyword systems.
4. Apply audience, geographic, language, device and scheduling controls appropriately.
5. Evaluate paid-search decisions using performance metrics, experimentation and governance principles.

1. Paid Search as a Market and Media System

A search advertisement is delivered at the intersection of four elements: a user’s query, the advertiser’s eligibility settings, the platform’s auction and the landing-page experience. The advertiser does not simply “buy a position.” Instead, the platform evaluates eligible ads whenever an auction is triggered. Eligibility may depend on budget, targeting, policy compliance, location, language, keyword or product relevance and account status.

Search advertising has two distinctive characteristics. First, it captures declared or inferred intent: a query such as “same-day laptop repair near me” may express stronger immediate demand than a broad demographic profile. Second, it is auction-based. The price and position can vary by query, competitor participation, device, time, location and predicted response probability.

The central managerial trade-off is between scale and control. Broad targeting can increase reach but also introduce irrelevant traffic. Narrow targeting can improve relevance but limit volume and create unstable learning. Effective management treats targeting as a set of testable hypotheses rather than permanent assumptions.

2. Platforms and Ad Formats

2.1 Major platform categories

General search platforms. Google Ads and Microsoft Advertising provide keyword-led advertising on search engines, partner properties and, depending on campaign type, additional networks. They support text ads, shopping formats, audience signals and automated bidding.

Retail and marketplace platforms. Amazon Ads and comparable retail-media systems place sponsored products, brands and display units close to purchase activity. Their data is especially valuable for product discovery and transaction-oriented measurement, although reporting and attribution are marketplace-specific.

Vertical and specialist search. Travel, employment, property, education and app-discovery platforms offer search-like inventory with specialized intent signals. Their auction logic may differ, but principles of relevance, bid control and measurement remain applicable.

2.2 Important ad formats

Responsive formats use platform systems to combine supplied assets. Automation can improve coverage, but it does not remove the advertiser’s responsibility for claims, brand safety, landing-page relevance or asset quality. A platform-generated combination should be evaluated as an advertising message, not treated as neutral software output.

3. Auction Fundamentals

3.1 Eligibility, ranking and price

In a simplified auction, the platform first identifies eligible ads. It then estimates their usefulness and determines whether they qualify for an impression. Modern platforms generally incorporate bid, predicted user response, relevance, landing-page experience, contextual factors and advertiser settings. The exact ranking formula is proprietary and varies by platform and format; therefore, no public equation should be treated as a complete representation of a live auction.

A useful conceptual model is:

This is a pedagogical approximation, not a guaranteed platform formula. “Bid” represents the advertiser’s bid or the value made available by an automated strategy. Quality or relevance factors may include expected click-through rate, ad relevance and landing-page experience. Contextual adjustments may include device, location, time, query meaning, competition and format. These factors are not necessarily multiplied in this exact manner in any platform.

In a generalized second-price-like system, an advertiser may pay the minimum amount needed to retain a position rather than the maximum bid. However, contemporary auctions often include reserve prices, quality signals, automated bidding and multiple placements. Consequently, the simple second-price explanation is useful for intuition but insufficient for forecasting actual cost.

3.2 Core performance metrics

The following metrics are widely used, although platform definitions and attribution windows must be checked before comparison.

CTR is click-through rate, expressed as a percentage. Clicks and impressions must refer to the same reporting scope and period.

CPC is average cost per click, expressed in the account’s currency per click. It is undefined when clicks equal zero.

CVR is conversion rate. The definition of “conversion” must be specified, because a form submission, qualified lead, purchase and page view are not equivalent outcomes.

CPA is cost per attributed conversion. It is undefined when conversions equal zero and can be misleading when conversion quality varies.

ROAS is return on advertising spend. Revenue must be defined consistently—for example, gross revenue, net revenue or contribution revenue—and should not be confused with profit.

A diagnostic identity links several metrics:

For example, if CPC is £2 and CVR is 5% (0.05), CPA is £2 ÷ 0.05 = £40. This relationship assumes that the same clicks, cost and conversions are used in all metrics and that conversion rate is not being reported under a different attribution rule.

4. Account and Campaign Architecture

4.1 Hierarchical design

A typical account hierarchy is:

Account → campaigns → ad groups → keywords and ads → landing pages.

Conversion actions, audiences, budgets and business data may operate across several levels. The account should reflect the organization’s billing, access-control and measurement boundaries. Campaigns should represent meaningful differences in objective, budget, geography, language, network, product category or bidding strategy. Ad groups should contain closely related themes and landing-page expectations.

The guiding principle is controlled homogeneity: items that require different decisions should be separated; items that can share the same decision should remain together. Excessive fragmentation produces small data sets and administrative burden. Excessive aggregation hides important differences and prevents budget control.

4.2 Campaign-level decisions

Campaign-level separation is usually justified when one or more of the following differs materially:

• budget or budget priority;
• geography or language;
• business objective;
• conversion goal;
• product margin or value;
• network or inventory type;
• bidding strategy;
• regulatory or brand-safety requirement.

For example, a national e-commerce campaign and a local emergency-service campaign should not automatically share a campaign merely because both use search keywords. Their geography, hours, conversion actions and commercial economics differ.

4.3 Ad-group design

An ad group should represent a compact intent theme. Keywords such as “running shoes,” “men’s running shoes” and “road running shoes” may share an ad group if one message and landing page serve them credibly. Queries requiring different claims or destinations should be separated. A useful test is: “Can one ad and one landing page answer the likely need behind every keyword in this group?” If not, split the group or revise the targeting.

Overly granular “single-keyword” structures are not automatically superior. They may increase control in a stable, high-volume account but can create thin data, duplicated negatives and slower experimentation. The appropriate level depends on volume, query diversity, product complexity and operational capacity.

4.4 Naming and documentation

A naming convention should encode decision-relevant information without becoming unreadable. For example:

UK | Search | Nonbrand | Running Shoes | Generic | tCPA

Documentation should record purpose, owner, geography, conversion goal, budget logic, launch date, tests, exclusions and policy constraints. Versioned change logs are particularly valuable when automated bidding or broad matching changes traffic composition.

5. Keyword Match Types and Query Control

Match types are targeting controls, not literal guarantees. Platform behavior evolves, and close variants, semantic interpretation and automated expansion mean that the actual query may differ from the exact keyword text.

Broad match

Broad matching can reach queries related to the keyword’s meaning, including variations, implied intent and context. It offers scale and can work effectively with strong conversion tracking, sufficient data, smart bidding and systematic query review. Its risk is semantic drift: the platform may identify demand that appears related linguistically but is commercially unsuitable.

Phrase match

Phrase match generally targets searches that include the meaning of the keyword in a specified order or close semantic relationship, subject to platform rules. It offers a middle position between reach and control. Advertisers should not assume that phrase match restricts every word to an unchanging sequence.

Exact match

Exact match is designed for highly controlled intent but may still include close variants and semantically equivalent searches. It is appropriate when query precision, budget protection or message specificity is important. Exact does not mean “only the identical character string.”

Match-type selection

Selection should follow risk and evidence. A new account with limited measurement may begin with more controlled themes, then expand after search-term and conversion data establish quality. A mature account may use broad matching for discovery and scale, while retaining separate controls for high-value terms, brand protection and regulated categories.

6. Negative Keywords and Exclusion Logic

Negative keywords prevent ads from serving for unsuitable queries. They are essential for protecting budget, preserving message relevance and separating intent. Typical negative themes include:

• employment and careers when the advertiser sells a service;
• free, template or tutorial when a paid solution is required;
• support or returns when the campaign is for acquisition;
• unrelated meanings of an ambiguous term;
• competitor terms where policy, strategy or economics make participation inappropriate.

Negatives should be organized into shared lists and campaign-specific exclusions. A shared list may exclude “jobs,” “salary” and “training” from all commercial acquisition campaigns. A campaign-specific list may prevent a premium product campaign from receiving queries for entry-level products.

Negative keywords also perform an architectural function. They prevent internal competition between campaigns with different budgets or objectives. For instance, a brand campaign can be protected from generic campaigns through carefully tested exclusions, while exceptions are documented rather than assumed.

Query reports should be reviewed for both irrelevant terms and valuable new themes. Adding every unusual query as a negative can suppress legitimate demand; adding no negatives allows waste to accumulate. Decisions should consider query intent, conversion quality, marginal cost and expected future value.

7. Audiences, Geography, Language and Scheduling

7.1 Audience controls

Audiences may be used in observation mode, where performance is measured without restricting reach, or in targeting mode, where eligibility is limited to selected users. Audience categories can include remarketing lists, customer lists, in-market segments, affinity groups and platform-defined predictive segments. Audience labels are probabilistic or behavioral constructs, not perfect descriptions of individuals.

A strong operating sequence is to begin with observation where risk is low, compare performance by audience, and apply restrictions only when evidence and strategy justify them. Customer-list use requires lawful collection, valid consent where applicable, secure handling and compliance with platform rules.

7.2 Geographic targeting

Geographic settings may target users physically located in an area, users showing interest in an area, or both, depending on platform configuration. The distinction matters. A hotel may wish to reach people planning to visit a city, whereas a plumber may primarily need people physically present nearby.

Geographic analysis should account for serviceability, travel time, local competition, language, legal restrictions and population density. Location reports should be checked after launch because selected settings do not guarantee perfect geographic precision.

7.3 Language and device

Language targeting interacts with user settings, query language, creative language and landing-page language. It should not be treated as a simple nationality proxy. Device performance may differ because of screen size, call behavior, connection quality and user context. Device adjustments or separate campaigns should be based on conversion quality rather than CTR alone.

7.4 Scheduling

Ad scheduling can reflect opening hours, staffing availability, inventory cycles and observed conversion patterns. A business that cannot answer calls overnight may reduce call-focused activity outside staffed hours. However, low overnight volume does not automatically mean poor performance: some users research at night and convert later. Scheduling decisions should therefore use assisted conversions, delayed conversions and lead-quality data where available.

8. Landing Pages, Conversion Measurement and Bidding

Paid search cannot be evaluated solely at the ad level. The landing page must deliver the promised information, load reliably, support accessibility and make the desired action clear. Message continuity—from query to ad to landing page—is a major component of relevance and user trust.

Conversion measurement should distinguish micro-conversions, such as product views or brochure downloads, from macro-conversions, such as purchases or qualified opportunities. Imported offline outcomes are especially important for lead-generation organizations. A platform may report a form submission, but the business needs to know whether that submission became a viable customer.

Automated bidding depends heavily on signal quality. If the conversion action includes low-value events, the system may optimize toward volume rather than commercial value. If data is sparse, optimization can be unstable. Bidding strategy should therefore follow measurement maturity, data volume, value consistency and business constraints—not fashion.

9. Worked Application: A Regional Legal-Service Campaign

The following is an illustrative case, not a report of a real company.

Background and problem

A regional law firm wants enquiries for employment-dispute consultations in three cities. Its existing account combines brand terms, informational searches, recruitment searches and all locations in one campaign. The firm receives many irrelevant clicks for jobs, legal training and free templates. Reported cost per lead is acceptable, but qualified-lead rate is poor.

Analysis

A query review reveals four distinct intents: brand navigation, immediate consultation, informational research and employment-seeking. The firm also discovers that phone enquiries are valuable during office hours, while online forms are more useful in the evening. Different cities have different lead quality and service capacity.

The original structure prevents separate budgets, schedules and messages. A single conversion action counts every form equally, even when the enquiry is outside the firm’s service scope.

Approach

The redesigned structure is:

Ad groups are organized around “unfair dismissal,” “redundancy advice” and “employment solicitor,” with separate pages where the legal promise differs. Negative lists exclude jobs, vacancies, salary, course, training, template and free. The firm uses observation audiences initially and separates location settings for physical presence from interest where appropriate.

The conversion design records form submission, phone call, consultation booking and qualified case acceptance as separate events. The last event is imported from the customer relationship management system, subject to privacy and data-governance requirements.

Outcome and interpretation

After a learning period, the firm observes fewer total clicks but a higher proportion of serviceable enquiries. This is a desirable outcome: reducing irrelevant volume can improve economic efficiency even when CTR or impression share declines. City-level budgets are adjusted using qualified-case value and capacity rather than lead count alone.

Learning points

1. Architecture should reflect differences in commercial decisions.
2. Negative keywords are a continuing governance process, not a one-time upload.
3. Lead quality requires offline feedback.
4. A lower-volume campaign can be more valuable when intent is stronger.
5. Local targeting and scheduling should be validated against operational reality.

10. Governance, Ethics and Contemporary Developments

Paid search involves privacy, fairness, transparency and consumer-protection obligations. Advertisers must avoid deceptive claims, unsupported guarantees, discriminatory targeting, hidden pricing and misleading landing pages. Regulated sectors may require special disclosures. Customer lists and remarketing data should be collected, stored and activated lawfully, with access controls and retention policies.

Automation and AI now assist with keyword expansion, asset generation, bidding, audience modeling and query interpretation. Their advantages include scale, rapid testing and pattern detection. Their limitations include opaque optimization, biased training signals, accidental policy violations, brand-inappropriate language and over-optimization for easily measured actions. Human governance should include approved claims, exclusion rules, monitoring thresholds, audit logs and clear accountability for generated assets.

Privacy changes, reduced third-party signals, consent requirements and modeled conversions make measurement less deterministic. Attribution reports should therefore be interpreted as estimates under defined rules, not as literal observations of every customer journey. Incrementality experiments, geo tests and controlled holdouts can complement platform-reported attribution when resources permit.

Best-Practice Checklist

• Begin with a documented business objective and primary conversion.
• Separate campaigns only when budget, geography, objective, value or control requirements differ.
• Keep ad groups thematically coherent and landing-page relevant.
• Treat match types as probabilistic controls that require query monitoring.
• Maintain shared and campaign-specific negative lists.
• Test geographic presence versus geographic interest settings.
• Use audience observation before applying restrictive targeting where appropriate.
• Align schedules with both user behavior and operational capacity.
• Import qualified offline outcomes for lead-generation accounts.
• Audit automated recommendations, generated assets and policy compliance.
• Evaluate profit, lead quality and incrementality—not CTR alone.

Chapter Summary

Paid search is an auction-based media system that connects query intent, advertiser eligibility, creative relevance, landing-page experience and bidding. Major formats include text, product, call, image, video and app advertising. Auction concepts such as bid, relevance, context and predicted response are useful for analysis, but platform ranking formulas are proprietary and dynamic.

Campaign architecture converts strategy into controllable units. Accounts contain campaigns, campaigns contain intent-based ad groups, and ad groups connect keywords, ads and landing pages. Match types manage reach and precision; negative keywords protect budgets and separate competing themes. Audiences, geography, language, devices and schedules should be selected according to evidence, serviceability and business operations. Measurement must distinguish clicks from valuable outcomes and should incorporate offline quality where possible. Automation can expand capability but requires human oversight, privacy compliance and rigorous testing.

Key Terms

Auction: A real-time process used to determine eligible ads, placement and often price.

Ad group: A collection of closely related keywords, ads and targeting themes.

Ad Rank: A platform-specific assessment influencing eligibility and placement.

Broad match: A match type designed to reach queries related to a keyword’s meaning.

Campaign: A management unit containing settings such as budget, geography and objective.

Conversion: A defined user action considered valuable by the advertiser.

Conversion rate: Conversions divided by clicks, usually expressed as a percentage.

Cost per acquisition: Advertising cost divided by attributed conversions.

Exact match: A controlled match type that can still include close variants or equivalent intent.

Landing page: The destination page reached after an advertisement is clicked.

Negative keyword: An exclusion that prevents ads from serving for specified query themes.

Phrase match: A match type intended to balance semantic reach and targeting control.

Quality or relevance signals: Platform assessments of expected usefulness and user experience.

ROAS: Revenue divided by advertising cost.

Search query: The user’s actual entered or interpreted search expression.

Scheduling: Controls that determine when ads may be eligible to serve.

Targeting: Rules defining which users, contexts or queries can receive an ad.

Value-based bidding: Optimization that uses differences in expected conversion value.

Review Questions

Short Answer Questions

1. Why is campaign architecture a strategic rather than purely administrative activity?
2. Distinguish a keyword from a search query.
3. State the formulas for CTR and CPA and define their variables.
4. What is the purpose of a negative-keyword list?
5. Why might a business use audience observation before audience targeting?

Long Answer Questions

1. Explain the main components of a paid-search auction and discuss why a simple bid-only model is inadequate.
2. Design an account structure for a multi-city retailer with different product margins and explain your separation decisions.
3. Compare broad, phrase and exact match, including their advantages, risks and appropriate use cases.
4. Discuss how landing-page relevance and conversion measurement influence automated bidding.
5. Evaluate the ethical and governance challenges created by AI-assisted paid-search management.

Analytical / Application Questions

1. An account spends £6,000, receives 3,000 clicks and records 150 conversions. Calculate CPC, CVR and CPA. Explain one limitation of interpreting these metrics without lead-quality data.
2. A campaign receives many searches for jobs, courses and free templates while advertising a paid professional service. Propose a negative-keyword governance process and explain how you would avoid suppressing valuable demand.
3. A local service operates in three cities, has different staffing capacity in each and receives both calls and online enquiries. Design its campaign, geographic and scheduling architecture, and identify the conversion data required for budget allocation.

Multiple Choice Questions with Answers

Question 1

Which account element most commonly controls budget and geographic settings?

A. Keyword

B. Campaign

C. Headline

D. Search query

Correct Answer: B. Campaign

Question 2

A campaign spends £500 and receives 250 clicks. What is its average CPC?

A. £0.50

B. £1.00

C. £2.00

D. £2.50

Correct Answer: C. £2.00

Question 3

Which metric is calculated as conversions divided by clicks?

A. CTR

B. CPC

C. CPA

D. CVR

Correct Answer: D. CVR

Question 4

What is the primary purpose of a negative keyword?

A. To increase the maximum bid automatically

B. To exclude unsuitable query themes

C. To create a new landing page

D. To change an ad into a shopping format

Correct Answer: B. To exclude unsuitable query themes

Question 5

Which statement about exact match is most accurate?

A. It always serves only the identical character string

B. It cannot use close variants

C. It is designed for control but may include close or equivalent intent

D. It automatically targets every query in the account

Correct Answer: C. It is designed for control but may include close or equivalent intent

Question 6

A firm separates campaigns by city mainly because the cities have different budgets and service capacity. Which architectural principle does this illustrate?

A. Separate units when important decisions differ

B. Place every keyword in one ad group

C. Use broad match exclusively

D. Optimize only for impressions

Correct Answer: A. Separate units when important decisions differ

Question 7

Why can a lower-volume campaign be commercially superior to a higher-volume campaign?

A. It always has a lower CPC

B. It may generate a higher proportion of qualified outcomes

C. It is exempt from auction dynamics

D. It guarantees higher impression share

Correct Answer: B. It may generate a higher proportion of qualified outcomes

Question 8

Which action is most appropriate before applying restrictive audience targeting in a new account?

A. Delete conversion tracking

B. Use observation where appropriate and compare performance

C. Exclude all returning users

D. Turn off geographic controls

Correct Answer: B. Use observation where appropriate and compare performance

Question 9

If CPC is £3 and conversion rate is 6% expressed as 0.06, what is CPA under the stated identity?

A. £0.18

B. £3.06

C. £18

D. £50

Correct Answer: D. £50

Question 10

Why should automated bidding be connected to qualified offline outcomes in lead generation?

A. To ensure the system optimizes toward business value rather than only raw form volume

B. To eliminate the need for landing pages

C. To guarantee every click becomes a sale

D. To prevent all broad-match traffic

Correct Answer: A. To ensure the system optimizes toward business value rather than only raw form volume