Media Buying, Budgeting and Advertising Economics

Introduction

Paid advertising converts financial resources into mediated exposure, interactions, installations, leads, customers and—ultimately—business value. Media buying is therefore more than purchasing impressions. It is the disciplined process of selecting audiences, inventory, platforms, bidding mechanisms, placements, schedules and measurement systems so that a defined communication objective is achieved at an economically acceptable cost.

Advertising economics connects campaign activity with managerial decisions. A marketer must be able to answer questions such as: How much does it cost to reach 1,000 people? Is a click economically valuable? Should the campaign optimise for traffic, conversions or profit? How much can be spent without exceeding the available budget? When does additional spending stop producing acceptable returns?

The calculations in this chapter assume that costs and outcomes are measured over the same period, attributed using a clearly specified method, and defined consistently. Reported platform metrics are estimates based on platform tracking and attribution; they are not automatically equivalent to finance-approved revenue or profit.

Learning objectives

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

• explain the media-buying process and major auction-based buying models;
• calculate and interpret CPM, CPC, CPV, CPI, CPA, CAC, ROAS and customer lifetime value;
• distinguish media efficiency metrics from business profitability metrics;
• select suitable bidding approaches for different campaign objectives;
• construct a budget, forecast results and monitor pacing;
• evaluate marginal returns and decide when spending should be shifted or stopped;
• recognise attribution, privacy, fraud and governance limitations in advertising measurement.

1. The media-buying system

A media plan specifies the audience, objective, message, channels, timing, placements, budget, buying method and success criteria. Media buying implements that plan by acquiring advertising opportunities. In digital environments, buying commonly occurs through real-time auctions: when a user loads an app or webpage, eligible advertisers compete for the available impression. The platform then selects an ad according to its auction rules, bid, estimated action probability, relevance or quality, and sometimes the advertiser’s total value to the platform.

The principal participants are:

• Advertiser: supplies the budget and commercial objective.
• Agency or trading team: plans, buys, optimises and reports media.
• Publisher or platform: supplies audience access and inventory.
• Ad exchange or supply-side platform: aggregates and sells inventory.
• Demand-side platform: helps buyers purchase inventory using targeting and bidding rules.
• Measurement and verification providers: estimate delivery, viewability, fraud, brand safety and outcomes.

A useful decision chain is:

Objective → audience → channel and inventory → buying model → bid and budget → measurement → optimisation.

The objective must precede the metric. Awareness campaigns may prioritise reach and completed views; performance campaigns may prioritise qualified leads, purchases or contribution margin. Optimising clicks when the business needs profitable customers can create impressive dashboard numbers but weak commercial results.

Common buying models

In auction systems, the advertiser may specify a manual bid, a maximum acceptable cost, or an automated goal. The amount actually paid can differ from the bid, particularly in second-price-like or value-based auctions. Consequently, a bid is not always the same as the final cost.

2. The advertising-funnel economics

A funnel links delivery to economic outcomes:

Impressions → views or clicks → visits → leads or installs → customers → revenue and contribution.

At every transition there is a rate. For example, click-through rate is clicks divided by impressions; conversion rate is conversions divided by the relevant clicks or sessions. Rates must be defined carefully: a platform may count link clicks, all clicks, landing-page views or unique clicks, and these are not interchangeable.

If a campaign generates 100,000 impressions, 2,000 clicks and 80 purchases, then:

• click-through rate = 2,000 ÷ 100,000 = 2%;
• click-to-purchase rate = 80 ÷ 2,000 = 4%;
• impression-to-purchase rate = 80 ÷ 100,000 = 0.08%.

The same campaign can therefore be described as strong or weak depending on the stage being examined. A useful report presents both the numerator and denominator, the time period, the attribution rule and the source of the data.

3. Core cost and return metrics

3.1 CPM: cost per mille

CPM expresses the cost of 1,000 impressions.

Spend is measured in currency and impressions are ad deliveries. The factor 1,000 converts cost per impression into cost per thousand impressions. If spend is $2,400 and delivery is 300,000 impressions:

CPM is appropriate for comparing the cost of inventory, but it does not establish that impressions were viewable, brand-safe, unique or effective. Reach and frequency should be examined alongside CPM. If 300,000 impressions reach 100,000 unique people, average frequency is 3 impressions per person.

3.2 CPC: cost per click

With $2,400 and 2,000 clicks, CPC is $1.20. CPC is useful for traffic acquisition, but inexpensive clicks can be undesirable if they have low intent or poor landing-page quality.

The relationship between CPM, click-through rate and CPC is:

Here CTR is expressed as a decimal. With an $8 CPM and a 2% CTR, CPC = 8 ÷ (1,000 × 0.02) = $0.40. This identity assumes CPM and CTR use the same impression and click definitions.

3.3 CPV: cost per view

A platform’s view may mean a minimum number of seconds played, a percentage of the video, or a completed view. Reports must state the operational definition. A $600 spend producing 30,000 qualifying views has a CPV of $0.02.

3.4 CPI: cost per install

If $5,000 produces 2,000 attributed installs, CPI is $2.50. CPI measures acquisition of an installed app, not necessarily a registered, active or paying user. Cohort retention and revenue are essential for evaluating app campaigns.

3.5 CPA: cost per action or acquisition

CPA is commonly calculated as:

The action must be specified: purchase, qualified lead, subscription, booking or another event. If $7,500 produces 150 qualified leads, CPA is $50 per lead. A lead CPA cannot be compared directly with a purchase CPA unless the action and downstream value are considered.

3.6 CAC: customer acquisition cost

Customer acquisition cost measures the cost of acquiring new customers, usually across a defined period:

A narrow paid-media CAC may include only campaign spend; a fully loaded CAC can include creative, agency, sales and technology costs. The chosen definition must be stated. If paid media costs $12,000, attributable sales support costs $3,000, and 300 new customers are acquired, fully loaded CAC is $50.

CAC is not necessarily equal to CPA. One customer may generate several conversions, and some attributed conversions may not become customers.

3.7 ROAS: return on advertising spend

ROAS is a ratio, often written as 4:1 or 400%. If attributed revenue is $40,000 and advertising spend is $10,000, ROAS = 4.0, or $4 revenue per $1 of advertising spend.

ROAS is not profit. If product cost, fulfilment, refunds and overhead consume most revenue, a high ROAS may still be unprofitable. A contribution-based view is often more useful:

If contribution margin is $18,000 on $10,000 spend, contribution ROAS is 1.8. The break-even ROAS depends on variable costs. If the contribution margin rate before advertising is 40%, break-even revenue ROAS is 1 ÷ 0.40 = 2.5, assuming no other relevant costs.

3.8 Customer lifetime value

Customer lifetime value (LTV or CLV) estimates the future economic value of a customer. A simple contribution-based model is:

This model assumes relatively stable behaviour and uses the same time unit for purchase frequency and expected lifetime. For example, an average order value of $60, four orders per year, a 45% contribution margin rate and an expected lifetime of 2.5 years produce:

A more rigorous discounted model is:

Here contribution_t is expected contribution in period t, r is the discount rate per period, and T is the forecast horizon. This model is appropriate when cash flows vary over time. Retention assumptions, discount rates and churn estimates create uncertainty; LTV should therefore be reported with a cohort definition and forecast horizon.

A practical acquisition rule is that expected LTV should exceed CAC by a sufficient margin. The required margin depends on cash-flow timing, risk, overhead and strategic objectives. LTV:CAC ratios are useful screening tools but should not replace contribution and payback analysis.

4. Budgeting and forecasting

A budget is a constraint and an investment plan. It should be allocated according to objective, expected marginal return, learning needs, audience size, seasonality and operational capacity—not simply divided equally among channels.

Bottom-up forecasting

A bottom-up forecast begins with planned spend and expected rates. For a click campaign:

Combining the equations:

Assume a $20,000 budget, $10 CPM, 1.5% CTR and 4% click-to-purchase conversion rate:

• impressions = ($20,000 ÷ $10) × 1,000 = 2,000,000;
• clicks = 2,000,000 × 0.015 = 30,000;
• purchases = 30,000 × 0.04 = 1,200;
• CPA = $20,000 ÷ 1,200 = $16.67.

This forecast is conditional, not certain. CPM, CTR and conversion rate can change with audience saturation, auction competition, creative fatigue, landing-page performance and seasonality.

Top-down forecasting

A top-down method starts with the required number of conversions and an allowable CPA:

If 800 purchases are required at an allowable CPA of $25, the planned budget is $20,000. This method is useful for financial planning, but the allowable CPA must be grounded in contribution economics rather than arbitrary platform benchmarks.

Pacing

Pacing compares actual spend with the amount that should have been spent by a given time. For a uniform daily budget:

A $30,000, 30-day campaign should have spent $10,000 after 10 days under linear pacing. If actual spend is $12,000, pacing is 120% of plan. This does not automatically mean performance is poor: an early test may intentionally front-load spending, or demand may be seasonal.

A useful pacing index is:

Values above 1 indicate overspending relative to the reference curve; values below 1 indicate underspending. Monitor daily spend, delivery, frequency, conversion lag and remaining budget together. Automatic platform budgets can accelerate unexpectedly when auctions change.

5. Bidding approaches and optimisation

Manual bidding

Manual bids provide direct control over bid ceilings and are useful when inventory value is well understood or strict cost control is required. They require more operational attention and can underdeliver if the bid is too low.

Automated conversion bidding

Automated systems adjust bids using signals such as device, context, audience and predicted conversion probability. Examples include target CPA, maximum conversions and target ROAS. These systems require sufficient conversion volume, stable event definitions and adequate learning time. A target is not a guarantee: if the target is unrealistic, delivery may fall sharply or the platform may seek low-quality conversions.

Value-based bidding

Value-based bidding assigns different values to outcomes. A purchase worth $200 should not necessarily be treated as equivalent to a $20 purchase. The organisation must send accurate values, exclude cancellations where possible, and protect the system from inflated or duplicated events.

Bidding objective selection

The lowest platform cost is not always the best result. Optimisation should move toward the deepest reliable event that can generate enough feedback for the algorithm.

6. Marginal returns and allocation decisions

Average performance describes the campaign as a whole; marginal performance describes the effect of the next unit of spending. This distinction is central to budget allocation.

For two spending levels:

Suppose spending rises from $10,000 to $14,000 and conversions rise from 500 to 650. The incremental 150 conversions cost $4,000:

If the original average CPA was $20, the additional spend is less efficient. It may still be acceptable if the allowable CPA is at least $26.67 and capacity exists.

In practice, response curves often show diminishing returns: early spending reaches high-propensity users, while later spending reaches less responsive users or repeatedly exposes the same audience. Budget should generally be shifted toward channels or segments with the strongest expected marginal contribution, not merely the strongest historical average ROAS.

Experiments improve allocation decisions. Holdout tests, geo experiments, conversion-lift studies and incrementality tests help estimate what would not have happened without advertising. Platform-attributed conversions can overstate incremental impact when users would have converted organically.

7. Measurement, attribution and data quality

Attribution assigns credit for a conversion to one or more marketing contacts. Common approaches include last-click, first-click, linear, position-based, time-decay and data-driven attribution. Each answers a different question. Last-click is simple but tends to reward lower-funnel contacts; first-click highlights discovery; multi-touch models distribute credit but depend on assumptions and observable data.

Three concepts must be separated:

1. Reported conversion: recorded by a platform or analytics system.
2. Attributed conversion: assigned credit under a selected rule.
3. Incremental conversion: caused by advertising compared with a credible counterfactual.

These quantities can differ substantially. Conversion windows also matter. A seven-day click window and a one-day click window will produce different results. Cross-device behaviour, consent restrictions, browser limitations, deleted cookies, offline sales and delayed conversions create measurement gaps.

Good measurement practice includes consistent event naming, deduplication, UTM governance, server-side or consent-aware measurement where lawful, CRM reconciliation, quality checks, and separation of platform reporting from finance reporting. Analysts should document currency, tax treatment, refunds, cancellation dates, time zones, attribution window and conversion definition.

8. Applied case: illustrative launch of a subscription app

Background

The fictional company StudyFlow launches a study-planning mobile application. It has a monthly subscription and wants to acquire paying users without exceeding a sustainable CAC.

Problem

The initial campaign reports a CPI of $3.00, which appears attractive. However, management does not know whether inexpensive installs produce active subscribers, whether the channel is incremental, or whether a $20,000 budget can be spent within 30 days.

Analysis

StudyFlow runs three channels with the following forecast assumptions:

Forecast installs are budget ÷ CPI. Forecast paid users are installs × install-to-paid rate.

• Social video: 3,000 installs and 240 paid users;
• Search: 1,400 installs and 210 paid users;
• Creators: 1,000 installs and 100 paid users.

Total expected paid users are 550. The blended acquisition cost is $20,000 ÷ 550 = $36.36. Estimated contribution LTV is:

• social video: 240 × $72 = $17,280;
• search: 210 × $90 = $18,900;
• creators: 100 × $80 = $8,000.

Total expected contribution is $44,180, producing contribution ROAS of $44,180 ÷ $20,000 = 2.21.

Approach and solution

The team does not judge the channels by CPI alone. Search has the highest CPI but the strongest downstream quality and an estimated CAC of $7,000 ÷ 210 = $33.33. Social video has CAC of $37.50, while creators have CAC of $40. Search therefore has the best expected acquisition economics, despite its higher install cost.

The team validates paid-user events, excludes duplicate installs, monitors seven-day activation and compares exposed regions with a holdout region. It also caps frequency in social video, refreshes fatigued creative and reserves part of the budget for testing rather than committing every dollar to the initial forecast.

Outcome

After the first week, search marginal CAC rises as high-intent inventory is exhausted, while a new social creative improves activation. The budget is reallocated using marginal paid-user contribution rather than the original CPI ranking. The final decision is based on measured paid subscriptions, contribution, retention and incrementality.

Learning points

The case demonstrates that an upstream metric such as CPI is not a sufficient measure of value. Downstream quality, LTV, attribution reliability and marginal performance determine whether a channel deserves additional investment.

9. Professional, ethical and regulatory considerations

Advertisers must comply with applicable privacy, consumer-protection, platform and sector regulations. Consent, lawful processing, purpose limitation, data minimisation and user rights are important when collecting identifiers or building audiences. Sensitive characteristics should not be exploited for discriminatory targeting. Children and vulnerable groups require particular protection.

Claims must be truthful, substantiated and presented without material deception. Frequency management reduces nuisance and potential harm. Brand-safety controls should address adjacency to harmful content, misinformation and illegal material. Advertising fraud—including bots, fake installs, click farms and domain spoofing—can inflate performance metrics and waste budgets. Independent verification, invalid-traffic filters and anomaly detection are useful safeguards.

AI is increasingly used for audience prediction, creative generation, bidding and budget recommendations. Human oversight remains necessary because automated systems can amplify biased data, optimise toward defective events, generate misleading creative or make decisions that are difficult to explain. A responsible workflow defines approved data, testing thresholds, audit trails, escalation procedures and a human decision owner.

10. Best-practice operating framework

A professional media-buying process should:

1. define the commercial objective and allowable economics;
2. specify conversion events and measurement ownership;
3. estimate funnel rates using historical or test data;
4. build base, optimistic and conservative forecasts;
5. assign budgets by expected marginal contribution and learning value;
6. select a bidding strategy compatible with data volume and objective;
7. establish pacing, frequency, quality and brand-safety controls;
8. reconcile platform results with analytics and finance records;
9. test incrementality where decisions are material;
10. document assumptions and update forecasts as evidence accumulates.

A dashboard should show spend, pacing, reach, frequency, CPM, CTR, CPC, conversion rate, CPA or CAC, revenue, contribution, ROAS, LTV, retention and marginal results. Metrics should be segmented by channel, audience, creative, device, geography and cohort when sample size permits.

Chapter Summary

Media buying is the planned acquisition of advertising inventory in support of a defined business objective. CPM, CPC, CPV, CPI and CPA describe costs at different stages of the funnel, while CAC links total acquisition cost to new customers. ROAS compares attributed revenue with advertising spend, but contribution ROAS is more closely related to profitability. LTV estimates the future contribution of a customer and must be grounded in explicit retention, frequency, margin and discounting assumptions.

Budget forecasts can be built from CPM, CTR and conversion rates or from target conversions and allowable CPA. Pacing controls delivery over time. Manual, automated and value-based bidding each have appropriate conditions, and automated optimisation depends on reliable conversion data. Average performance must be distinguished from marginal performance because additional spending commonly produces diminishing returns. Attribution is not the same as incrementality, and professional decisions require data governance, privacy protection, fraud controls, ethical targeting and reconciliation with finance.

Key Terms

CPM: Cost of one thousand ad impressions.

CPC: Cost paid or calculated for each click.

CPV: Cost per defined video view.

CPI: Cost per attributed app install.

CPA: Cost per defined conversion or action.

CAC: Cost of acquiring a new customer.

ROAS: Attributed revenue divided by advertising spend.

Contribution margin: Revenue remaining after specified variable costs.

LTV: Expected customer contribution over a defined lifetime or forecast horizon.

CTR: Clicks divided by impressions.

Conversion rate: Conversions divided by a specified opportunity base.

Frequency: Average number of impressions per reached person.

Pacing: Control of spend delivery against a time-based plan.

Attribution: Assignment of conversion credit to marketing contacts.

Incrementality: The additional outcome caused by advertising.

Marginal CPA: Cost of the additional conversions generated by extra spend.

Value-based bidding: Bidding that accounts for different economic values of outcomes.

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

Invalid traffic: Non-genuine or non-human activity that can distort delivery or results.

Review Questions

Short Answer Questions

1. What does CPM measure, and why is it insufficient as a complete effectiveness metric?
2. Distinguish CPA from CAC.
3. Why can a campaign with a high ROAS still be unprofitable?
4. What is meant by pacing in media buying?
5. Distinguish an attributed conversion from an incremental conversion.

Long Answer Questions

1. Explain the advertising-funnel economics from impressions to customer lifetime value, including the role of denominators in metric interpretation.
2. Compare manual, automated conversion and value-based bidding, identifying suitable conditions and risks for each.
3. Explain how a marketer can construct and critically evaluate a bottom-up budget forecast.
4. Discuss why LTV and CAC should be calculated using consistent cohort, cost and contribution definitions.
5. Evaluate the ethical, privacy and fraud-related risks involved in data-driven media buying.

Analytical / Application Questions

1. A campaign spends $12,000, receives a $6 CPM, achieves a 1.25% CTR and converts 3% of clicks. Calculate impressions, clicks, conversions and CPA. State the assumptions used.
2. A channel’s spend increases from $8,000 to $11,000, while attributed revenue increases from $24,000 to $31,500. Calculate average ROAS at both levels and marginal ROAS for the additional spend. Which figure should guide the next budget decision, and why?
3. An app campaign reports a CPI of $2.50, but only 6% of installers become paying users. If the contribution LTV of a paying user is $55, calculate the expected contribution per install and assess whether the campaign appears sustainable before considering overhead and incrementality.

Multiple Choice Questions with Answers

Question 1

A campaign spends $2,400 and delivers 300,000 impressions. What is its CPM?

A. $0.80

B. $8.00

C. $80.00

D. $800.00

Correct Answer: B. $8.00

Question 2

Which metric is calculated as advertising spend divided by attributed new customers?

A. CPC

B. CPM

C. CAC

D. ROAS

Correct Answer: C. CAC

Question 3

A campaign has a $10 CPM, a 2% CTR and consistent impression and click definitions. What is its CPC?

A. $0.05

B. $0.20

C. $0.50

D. $2.00

Correct Answer: C. $0.50

Question 4

A campaign spends $10,000 and generates $40,000 in attributed revenue. What is its ROAS?

A. 0.25

B. 2.5

C. 4.0

D. 40.0

Correct Answer: C. 4.0

Question 5

Which statement best distinguishes ROAS from profitability?

A. ROAS always includes product and fulfilment costs.

B. ROAS compares attributed revenue with ad spend, so it is not automatically profit.

C. ROAS measures only unique reach.

D. ROAS is identical to customer lifetime value.

Correct Answer: B. ROAS compares attributed revenue with ad spend, so it is not automatically profit.

Question 6

A $20,000 campaign has a $10 CPM, 1.5% CTR and 4% click-to-purchase rate. How many purchases are forecast?

A. 120

B. 600

C. 1,200

D. 2,000

Correct Answer: C. 1,200

Question 7

A campaign’s spend rises by $4,000 and conversions rise by 150. What is the marginal CPA?

A. $6.67

B. $20.00

C. $26.67

D. $60.00

Correct Answer: C. $26.67

Question 8

Which bidding approach is most directly designed to account for different economic values of conversions?

A. Value-based bidding

B. Uniform frequency capping

C. Fixed sponsorship buying

D. Reach-only buying

Correct Answer: A. Value-based bidding

Question 9

If average order value is $60, annual purchase frequency is 4, contribution margin rate is 45% and expected lifetime is 2.5 years, what is simple contribution LTV?

A. $108

B. $270

C. $600

D. $675

Correct Answer: B. $270

Question 10

Which result most directly estimates what advertising caused rather than merely what it received credit for?

A. Last-click attribution

B. Platform-reported impressions

C. Incrementality testing

D. Average frequency

Correct Answer: C. Incrementality testing