Every year, companies around the world spend over one trillion dollars trying to get people to notice things. They buy seconds of time on television, pixels on web pages, panels on highway overpasses, space inside podcasts, and milliseconds of attention on social media feeds. The industry that mediates all of this - buying, selling, measuring, and optimizing attention at industrial scale - is one of the most complex, data-intensive, and economically significant systems ever constructed.
And yet most people experience advertising only from the outside: as something that interrupts what they were trying to watch, read, or listen to. The machinery behind that interruption - the auctions happening in 100 milliseconds, the audience segments built from billions of data points, the television ratings that determine what gets made and what gets cancelled, the algorithm deciding which billboard message to show based on traffic patterns - is almost entirely invisible to the people it targets.
This guide makes it visible. We will cover every major advertising medium in depth - television, digital display, search, social media, out-of-home, radio, podcasts, connected TV, and print - explaining exactly how each one works, how money flows through it, what advertisers pay and why, and how the industry is changing under the pressures of data, technology, and shifting consumer behavior.
01 - FoundationsThe Core Mechanics of Advertising
Advertising, at its most abstract, is the paid placement of a message in front of an audience. The value of that placement depends on two factors: how many people see it (reach) and how relevant those people are to the advertiser (targeting). Every innovation in advertising over the past century has been an attempt to improve one or both of those factors.
Early newspaper advertising in the 1800s had enormous reach but almost no targeting - anyone who bought the paper saw the ad. Television in the 1950s added visual and audio storytelling but targeting was still crude (demographic assumptions based on what kind of show someone was watching). The internet fundamentally changed the equation by making it possible to target based on actual behavior - what people searched for, what they bought, where they went - rather than demographic proxies.
The three-party structure
Almost every advertising transaction involves three parties, not two:
- The advertiser (brand, business, political campaign) - who wants to deliver a message and has a budget to spend
- The publisher (TV network, website, billboard company, radio station) - who owns the audience's attention and sells access to it
- The audience (viewers, readers, listeners, scrollers) - who provide their attention in exchange for content, typically without paying money directly
This structure - audiences subsidizing content they consume by also consuming ads - is the business model of most of the world's media. The question of where precisely the money flows within and between these three parties is what makes advertising economics so complex.
Reach, frequency, and effective coverage
Every media plan - whether for a local pizza shop or a global consumer brand - is built around a few fundamental metrics. Reach is the total number of unique people exposed to an ad at least once. Frequency is how many times the average exposed person sees it. The product of reach and frequency gives gross rating points (GRPs) in traditional media, or impressions in digital - the raw mass of exposure delivered.
Advertisers generally target an "effective frequency" - the minimum number of times a person needs to see an ad before it influences their behavior. Too few exposures and the message doesn't register. Too many and the audience tunes it out (or becomes actively annoyed). Finding the optimal frequency is one of the central optimization problems in media planning.
02 - TelevisionBroadcast, Cable & the Upfront Market
Television remains the single most powerful medium for building brand awareness at scale, despite decades of predictions about its decline. In 2024, US television (broadcast + cable) still commanded over $60 billion in advertising revenue. The reason is simple: nothing else can put your message in front of 10 million people simultaneously with full audio, video, and emotional storytelling.
How TV ad time is sold: The Upfront Market
The television advertising market operates on a peculiar annual ritual called the Upfront Market. Each spring, broadcast and cable networks present their upcoming season's programming to advertisers and media buyers. Advertisers then commit - "upfront," months before the content airs - to purchasing large volumes of ad time at negotiated rates. In exchange, they get pricing stability, guaranteed inventory in the most desirable shows, and a degree of protection against rate increases.
The networks love the upfront system because it allows them to lock in revenue before a single episode airs. The risk is on the advertiser: they're buying based on projected ratings, and if a show underperforms, the network owes them "make-goods" - additional spots to compensate for the shortfall. The upfront market typically accounts for 70-80% of all network prime-time inventory. The remaining inventory is sold in the Scatter Market - closer to air date, at higher and more volatile prices that fluctuate with demand.
Ratings, GRPs, and CPMs
Television advertising is bought and sold in currency of ratings. A rating point represents 1% of the total potential audience - in the US context, a 1.0 rating for adults 18-49 means 1% of all adults aged 18-49, roughly 1.3 million people. Advertisers buy GRPs (Gross Rating Points) - a schedule of spots adds up to a total GRP number that represents the aggregate weight of a campaign.
The price is expressed as CPM (Cost Per Thousand impressions) - the cost to reach 1,000 viewers in the target demographic. Prime-time network CPMs for the coveted Adults 18-49 demographic typically range from $30 to $60 per thousand in normal market conditions. The Super Bowl is an extreme outlier: at $9+ million per 30 seconds reaching 110+ million viewers, the CPM is actually quite competitive - roughly $80 per thousand - but the absolute cost puts it out of reach for all but the largest global brands.
Nielsen and audience measurement
The currency of TV advertising has historically been Nielsen ratings - gathered via a panel of approximately 40,000 US households with metering devices that track what's on the television set, combined with people meters that log who is in the room. Nielsen extrapolates from this panel to the full national audience. This system, designed in an era of three broadcast networks, is now showing serious strain in a world of streaming, time-shifting, and multi-screen viewing.
The industry is actively transitioning toward alternative measurement currencies - including iSpot.tv, VideoAmp, and Comscore - that incorporate streaming data, return-path data from smart TVs and cable boxes (which capture viewing more completely than a panel), and digital ad verification signals. This "currency wars" moment in TV measurement is reshaping how advertising deals are negotiated.
How local TV advertising works
National network buys are only one layer of the TV advertising system. Local television - the ABC, NBC, CBS, and Fox affiliates in each market - sells its own advertising within local newscasts, daytime programming, and during breaks in network programming that the network assigns to affiliates. Local TV ad time is far cheaper than national (a 30-second prime-time spot in a mid-size market might cost $500 to $3,000) and is the medium of choice for car dealerships, personal injury law firms, local retailers, and political campaigns.
Political advertising is one of the most significant drivers of local TV revenue in the US. In presidential election years, political ad spending can exceed $8-10 billion nationally, with the majority flowing to local TV stations in battleground states. Federal law mandates that stations must offer political candidates the lowest unit rate - the best rate they charge any advertiser for the same time slot - and cannot reject federal candidate ads even if the content is objectionable.
03 - DigitalDisplay, Programmatic & Real-Time Bidding
When you load a web page and an ad appears, a remarkably sophisticated sequence of events has occurred in the roughly 100 milliseconds between your click and the page rendering. An auction has been conducted, won, and executed. Your profile has been assembled from dozens of data sources, matched to an advertiser's target audience, competed for by multiple bidders, and a winner has delivered a creative to fill the space - all before you could blink.
This is programmatic advertising - the automated buying and selling of digital ad inventory using software, data, and real-time auctions rather than direct human negotiation. Programmatic now accounts for approximately 90% of all digital display advertising in developed markets.
The programmatic stack: who does what
A website or app integrates with a Supply-Side Platform (SSP) - a platform that manages the publisher's available ad inventory and makes it available to buyers through ad exchanges. The publisher sets a floor price (minimum CPM they'll accept) and controls which categories of advertiser can appear.
The moment a user loads a page with an ad slot, the SSP sends a bid request to multiple ad exchanges and demand-side platforms. This request includes information about the ad slot, the page content, and - critically - data about the user: device type, browser, location, and pseudonymous identifiers that link to audience profile data.
Advertisers buy digital inventory through Demand-Side Platforms (DSPs) - Google DV360, The Trade Desk, Amazon DSP, and others. The DSP receives the bid request, checks whether the user matches any of the advertiser's target audience segments, and if so, submits a bid in real time.
The ad exchange runs a real-time auction among all submitted bids. In a second-price auction (historically standard), the winner pays $0.01 above the second-highest bid. First-price auctions (now more common) require the winner to pay their full bid. The entire process completes in under 100 milliseconds.
The winning advertiser's ad is delivered. Tracking pixels log the impression, and if the user later clicks or converts, that action is attributed to the ad. The data feeds back into the DSP's optimization algorithm to inform future bids.
The cookie and the death of third-party tracking
The entire personalization architecture of programmatic advertising historically rested on third-party cookies - small identifiers stored in your browser by ad networks (not by the website you were visiting). These cookies allowed ad tech companies to recognize you across different websites and build behavioral profiles: you visited a car website, so car ads followed you to the news site, the sports site, and the recipe site you visited next. This is called retargeting.
The deprecation of third-party cookies - by Firefox and Safari already, and eventually by Chrome (after multiple delays) - is the most significant structural change to happen to digital advertising in a decade. Without cross-site tracking, the precise behavioral targeting that defined programmatic advertising must be replaced with contextual targeting (ads matched to page content rather than user history), first-party data (information advertisers collect directly from their own customers), and privacy-preserving techniques like Google's Privacy Sandbox. The industry is in the middle of this transition, with considerable uncertainty about which approaches will dominate.
Ad fraud: the industry's dirty secret
Digital advertising has a significant and persistent fraud problem. Invalid traffic (IVT) - fake impressions generated by bots rather than real humans - is estimated to cost advertisers $100 billion+ annually by 2024. Fraud takes many forms: bot networks that simulate human browsing to generate fake impressions, domain spoofing (a fraudulent site pretending to be a premium publisher to command higher CPMs), ad stacking (multiple ads loaded invisibly on top of each other so only the top one is visible but all are "counted"), and pixel stuffing (ads rendered at 1x1 pixel, invisible to humans but counted as an impression).
The industry has responded with verification tools (DoubleVerify, IAS - Integral Ad Science, MOAT) that apply filters to block invalid traffic and certify that ads were actually viewable by a human on the claimed domain. The IAB's ads.txt and sellers.json protocols allow publishers to publicly declare which companies are authorized to sell their inventory - reducing domain spoofing. But fraud remains a persistent cost center in digital advertising, and advertisers who do not actively implement brand safety and fraud prevention measures continue to waste significant percentages of their budgets.
The most sophisticated media operation in human history turned out to be largely buying bot traffic and calling it reach. Marc Pritchard, Chief Brand Officer, Procter & Gamble, 2017 IAB Annual Leadership Meeting
04 - SearchGoogle Ads & the Pay-Per-Click Model
Search advertising is unique among all advertising media: it is the only major channel where the user initiates the interaction. Someone typing "best running shoes for flat feet" is not an interrupted bystander - they are an active, self-declared prospect. This is why search advertising commands dramatically higher prices than display (CPCs of $1 to $50+ versus display CPMs of $1 to $10) and why it generates the highest and most measurable return on investment of any digital channel.
How Google's auction works
Google Search ads operate through a real-time auction that runs every time someone enters a search query. But it is not simply the highest bidder who wins the top position. Google determines ad rank using a formula that combines the advertiser's maximum bid (the most they're willing to pay per click) with a Quality Score - a rating of 1 to 10 that reflects expected clickthrough rate, ad relevance to the query, and the quality of the landing page experience.
The effect of Quality Score is profound. An advertiser with a Quality Score of 10 bidding $1.00 per click can outrank an advertiser with a Quality Score of 3 bidding $3.00 per click. This means that better-crafted, more relevant ads with excellent landing pages can win premium positions at lower costs - rewarding genuine relevance over raw spending power. The price paid per click is determined by a modified second-price auction: you pay the minimum amount needed to maintain your position, which is typically somewhat below your maximum bid.
Keyword economics
The price of a keyword is determined by the competition among advertisers targeting it. Keywords that signal high commercial intent with large addressable markets are the most expensive. The most costly keywords in Google Ads include terms related to legal services ("mesothelioma lawyer" at $900+ per click), financial products ("refinance mortgage"), insurance ("car insurance quotes"), and medical services ("drug rehabilitation"). Meanwhile, keywords in less competitive niches with lower customer lifetime values may cost $0.10 to $0.50 per click.
Shopping ads, Performance Max, and automation
Beyond text ads, Google offers Shopping Ads (product listings with images, prices, and ratings that appear at the top of search results for product queries), Local Service Ads (lead-generation ads for local service businesses with a pay-per-lead model), and the increasingly dominant Performance Max campaigns - an AI-driven format where advertisers provide creative assets and audience signals, and Google's algorithm automatically places ads across Search, Display, YouTube, Gmail, and Maps to maximize conversions.
Performance Max represents Google's push toward full automation: the advertiser cedes control over placement, bidding, and to a significant degree targeting, in exchange for (claimed) better performance through machine learning. This tension between control and automation defines the current state of search advertising, with sophisticated advertisers debating how much to trust the "black box" of automated campaigns versus maintaining manual oversight.
06 - Out-of-HomeBillboards, Transit & Digital OOH
Out-of-home advertising - billboards, bus shelters, transit stations, airport signage, building wraps, stadium displays, and the growing ecosystem of digital screens in retail environments - is arguably the most honest form of advertising. It makes no pretense of being anything other than an ad. It sits in the physical world, demanding attention through scale, placement, and creative impact rather than through personalization algorithms.
Traditional static billboards
A traditional static billboard is a straightforward transaction: an advertiser rents a physical panel for a defined period (typically 4-week cycles), a printer produces the artwork (vinyl or paper), and a contractor installs it. Pricing is based on the billboard's DEC (Daily Effective Circulation) - the estimated number of people who pass it daily, derived from traffic counts and audience research. A billboard on a major urban freeway might have a DEC of 200,000 to 500,000 vehicles. The CPM for traditional billboards is typically $2 to $8 - far below digital media - but because the exposure is purely passive (no one "chooses" to see a billboard), frequency builds naturally for commuters who pass the same route daily.
Pricing varies enormously by location: a billboard in Times Square, New York costs over $50,000 per month. A comparable-size board on a secondary road in a mid-size city might be $1,500 per month. The primary operators of US billboard inventory are Lamar Advertising, Clear Channel Outdoor, and Outfront Media - together controlling the vast majority of premium inventory.
Digital OOH (DOOH) - the transformation
Digital Out-of-Home represents the fastest-growing segment of the OOH market. Digital screens allow multiple advertisers to share the same physical location (rotating ads every 8-10 seconds), eliminate production costs, enable dynamic creative that changes based on time, weather, or events, and increasingly support programmatic buying through the same DSP infrastructure used for online advertising.
A digital billboard running 6 advertisers per rotation has six times the revenue potential of an equivalent static board, which explains why operators have invested heavily in converting inventory. Modern DOOH has features that would have seemed extraordinary even a decade ago: screens that change their message based on current weather (a hot day triggers cold drink advertising, rain triggers umbrella retailers), traffic data (messages adjusted for traffic speeds - slower traffic allows more reading time, so longer copy is served), and even mobile device aggregation data that characterizes the audience profile of people who pass a given location throughout the day.
Programmatic DOOH
The integration of DOOH into programmatic buying platforms is enabling true audience-based purchasing of physical world advertising. An advertiser targeting adults 25-45 interested in health and fitness can now buy OOH impressions near gyms during morning commute hours, in the same campaign workflow used to buy Instagram Stories and Google display ads. While the targeting is based on aggregated location data rather than individual profiles (preserving privacy), the ability to align digital and physical media buying with consistent audience logic is a significant development for multi-channel campaign planning.
Transit advertising
Transit advertising - on bus exteriors, inside subway cars, at train stations and bus stops - offers a unique environment: captive audiences with significant dwell time and minimal competing media stimuli. A subway passenger on a 30-minute commute has extended, repeat exposure to station platform ads and interior car cards. Transit advertising in major cities (New York's MTA, London Underground, Paris Metro) commands significant premiums and is often used for entertainment launches, fashion brands, and political campaigns targeting urban demographic profiles.
07 - RadioTerrestrial, Satellite & Streaming Audio
Radio is one of the most persistently misunderstood media channels. Despite the rise of streaming music and podcasts, terrestrial radio reaches approximately 273 million Americans (89% of adults) every week - more than any other single medium. The reason is the car: radio remains the default audio environment for the approximately 90% of Americans who commute by personal vehicle, providing an involuntary but undivided audience during the high-attention morning and evening drive-time windows.
Radio advertising is sold in 30-second and 60-second units, priced using a combination of Average Quarter-Hour (AQH) audience (how many people listen during the average 15-minute period), CUME (total unique listeners in a week), and daypart. Morning Drive (6am-10am) and Afternoon Drive (3pm-7pm) command significant premiums over midday and evening spots because of the larger, more attentive audiences.
Production costs for radio are minimal - a well-written script, a voice talent, and some background music is all that's needed. This low barrier makes radio accessible to small local businesses in a way that television production costs prohibit. A small business can run a credible radio campaign in a local market for as little as $2,000 to $5,000 per month including production.
Satellite radio (Sirius XM in the US) operates on a subscription model that includes ad-supported tiers on some channels. Digital audio streaming platforms like Spotify and Pandora have created new audio advertising inventory at scale, with the added benefit of digital targeting capabilities applied to audio ads - a significant evolution beyond traditional terrestrial radio's demographic-based buying.
08 - PodcastThe Most Trusted Ad Format in Media
Podcast advertising has grown from a niche experiment to a $4+ billion US market in under a decade, driven by a combination of explosive audience growth, uniquely intimate listener relationships, and ad recall rates that far exceed any other media channel. A 2023 Nielsen study found that 71% of podcast listeners say they are more aware of brands after hearing them advertised on a podcast - compared to 55% for TV ads.
The reasons trace back to the fundamental nature of the medium. Podcast listeners choose specific shows, typically listen to full episodes, and often develop parasocial relationships with hosts they've listened to for years. When that trusted host reads an ad - especially in a conversational, personalized style that departs from a scripted pitch - the recommendation carries genuine credibility that a pre-roll ad on a website simply cannot approximate.
Ad formats: pre-roll, mid-roll, host-read
Podcast ads are typically placed at three positions: pre-roll (before content starts - typically 15-30 seconds, highest skip rate), mid-roll (inserted during the episode - typically 60-90 seconds, lowest skip rate, highest engagement and premium pricing), and post-roll (after episode ends - lowest audience retention, lowest cost). Mid-roll host-read ads are the gold standard of podcast advertising, commanding CPMs of $20 to $50 for mainstream shows and $60 to $150+ for premium business, finance, or true crime podcasts with highly sought-after audiences.
Baked-in vs. dynamic insertion
Podcast ads can be either baked-in (permanently embedded in the audio file, heard by all future listeners) or dynamically inserted (served through ad insertion technology at the time of download, allowing different ads to be served to different listeners and campaigns to be switched out). Baked-in ads are preferred by direct-response advertisers who want the credibility of a genuine host endorsement; dynamic insertion allows for precise targeting, campaign management, and the kind of frequency capping that direct-response performance campaigns require. Most large podcast networks now use a hybrid approach.
Measurement: the download problem
Podcast measurement remains the medium's most significant commercial limitation. Unlike streaming video (where server-side data captures exact viewing) or display advertising (with pixel-level tracking), podcast consumption is largely measured through downloads - which indicate the file was requested but not that it was actually listened to, or listened to at all the way through. Attribution - knowing that a listener who heard a promo code on a podcast actually converted to a customer - relies on imperfect proxies: unique promo codes, custom landing page URLs, or survey-based brand lift measurement. Spotify's fully streamed podcast ecosystem provides better data than the open podcast market, which is one significant advantage the platform holds for advertisers who want measurement confidence.
09 - CTVConnected TV & Streaming: The New Primetime
Connected TV represents the most significant structural shift in television advertising in decades. As audiences migrate from linear broadcast and cable to streaming services, the inventory they watch is moving from the traditional TV buying ecosystem (Upfronts, GRPs, Nielsen) to a digital advertising ecosystem (programmatic buying, DSPs, digital attribution). CTV combines the creative power and premium environment of television content with the targeting precision and measurability of digital advertising.
The CTV advertising market is dominated by a mix of pure-play streaming platforms (Hulu, Peacock, Paramount+, Max, Disney+), ad-supported streaming services (Tubi, Pluto TV, Roku Channel), and the major streaming platforms that have introduced ad-supported tiers (Netflix, Amazon Prime Video). Netflix launched its ad-supported tier in November 2022; Amazon forced its Prime Video subscribers into ads in early 2024 (with an opt-out for a higher subscription price) - together bringing over 300 million potential ad-supported streaming viewers into the CTV advertising market almost overnight.
Why CTV CPMs are higher than linear TV
CTV impressions command CPMs of $25 to $65 - typically 2 to 3 times higher than equivalent linear TV CPMs. The premium reflects several advantages: audience targeting based on household-level demographics and viewing behavior rather than panel-based demographic proxies, ad completion rates near 95% (users cannot skip most CTV ads), frequency capping that prevents the same household from seeing the same ad 15 times (a common problem in linear TV), and attribution capabilities that can track whether a CTV ad viewer later visited a website, made a purchase, or visited a physical store. For advertisers, the measurability premium alone justifies the higher CPM.
AVOD, SVOD, FAST: the streaming alphabet
- SVOD (Subscription Video on Demand) - pure subscription, no ads (Netflix original model, Disney+ launch model)
- AVOD (Ad-Supported Video on Demand) - free to consumers, funded by advertising (Tubi, Peacock free tier)
- FAST (Free Ad-Supported Streaming TV) - linear-style channels on streaming platforms, ad-supported (Pluto TV, Samsung TV Plus)
- Hybrid (SVOD + AVOD) - paid subscription options with and without ads (Netflix, Hulu, Disney+, Max)
10 - PrintNewspapers & Magazines in the Digital Era
Print advertising - newspapers and magazines - is the only major advertising medium in structural secular decline. US newspaper ad revenue has fallen from $49 billion in 2005 to under $9 billion in 2023. The cause is the internet's destruction of the classified advertising business (once the dominant profit center for local newspapers) and the migration of display advertising to digital channels with superior targeting and measurability.
Magazine advertising has fared somewhat better in premium segments. Luxury publications (Vogue, The Economist, Architectural Digest), specialist interest titles (hunting, cooking, automotive), and professional/trade publications retain meaningful advertising revenue because they reach high-affinity, often affluent audiences who are difficult to target efficiently through digital channels. A full-page four-color ad in Vogue still commands $160,000+ and carries a prestige signal that an Instagram ad simply cannot replicate.
Print ads are sold by page size (full page, half page, quarter page, spread) and by position (back cover commands the highest premium, followed by inside front cover, then inside back cover, with run-of-book being lowest). Unlike digital advertising, print has physical permanence - a magazine may be read multiple times over months, and pass-along readership (copies shared between multiple readers) means actual readership can be 3-5x circulation numbers.
Pricing & Revenue Models
11 - PricingCPM, CPC, CPA & Every Revenue Model
Advertising pricing models have proliferated dramatically with the rise of digital media. Understanding the differences between them is critical to evaluating the true economics of any advertising investment.
Why the pricing model matters as much as the price
The choice of pricing model fundamentally allocates risk between advertiser and publisher. A CPM deal puts the performance risk entirely on the advertiser - they pay regardless of whether anyone clicks or buys. A CPA deal puts performance risk on the publisher or network - they only get paid if the user converts, incentivizing them to deliver quality traffic rather than just volume. A CPC model sits in between.
In general, publishers prefer CPM deals (predictable revenue regardless of user behavior). Advertisers prefer CPA or CPC deals (costs are linked to actual performance). The prevalence of CPM pricing in most premium digital advertising reflects the power of platform publishers - it takes a very sophisticated advertiser, or an affiliate network context, to achieve pure CPA pricing for mainstream display advertising.
| Medium | Pricing Model | Typical CPM Range | Targeting | Measurability | Min. Budget |
|---|---|---|---|---|---|
| Network TV (prime) | CPM (GRPs) | $30-$60 | Low | Indirect | $250K+ |
| Cable TV | CPM (GRPs) | $10-$30 | Medium | Indirect | $50K+ |
| CTV/Streaming | CPM | $25-$65 | High | Good | $10K+ |
| Google Search | CPC | N/A ($0.10-$900 CPC) | Very High | Excellent | $500+ |
| Google Display | CPM / CPC | $1-$5 | High | Good | $500+ |
| Meta (Facebook/IG) | CPM / oCPM | $8-$25 | Very High | Good | $500+ |
| YouTube | CPV / CPM | $10-$30 | High | Good | $1K+ |
| TikTok | CPM / CPC | $10-$30 | High | Fair | $50+ |
| Podcast (mid-roll) | CPM | $20-$150 | Medium | Weak | $5K+ |
| Radio | CPM / spot rate | $5-$15 | Medium | Indirect | $2K+ |
| Billboard (static) | Monthly flat | $2-$8 | Low | Weak | $1.5K/mo |
| Digital OOH | CPM / weekly | $3-$12 | Medium | Improving | $5K+ |
| Print (magazine) | Per-insertion | $10-$40 | Low | Very Weak | $10K+ |
| Programmatic display | CPM | $0.50-$5 | Very High | Excellent | $500+ |
12 - Ad TechThe Ecosystem Behind Every Digital Ad
The digital advertising industry has spawned an extraordinarily complex ecosystem of technology companies - often called "ad tech" - that mediate between publishers and advertisers. Understanding this ecosystem helps explain why publishers often receive only 30-50 cents of every dollar advertisers spend.
The tax of the middle: where advertiser dollars go
One of the most controversial topics in digital advertising is the "ad tech tax" - the percentage of every advertiser dollar consumed by intermediaries before it reaches the publisher. A 2020 ISBA (UK) study tracked £1 of programmatic spend through the supply chain and found that publishers received only 51 pence. The remaining 49% was consumed by various intermediary tech fees. A subsequent ANA (US) study in 2023 found that 29% of programmatic spend was "unresolvable" - impossible to trace to a final publisher, likely consumed by intermediaries or fraud.
This opacity is being addressed through initiatives like sellers.json, SupplyChain objects (which create an auditable chain of custody from publisher to buyer), and the shift toward direct buying - advertisers working directly with publishers through Preferred Deals and Private Marketplace deals that bypass the open exchange and its associated intermediary costs. For large advertisers, investing in in-house programmatic capability (operating their own trading desk with a DSP rather than paying agency fees on top of tech fees) can significantly reduce the total cost of programmatic buying.
13 - RevenueHow the Money Flows Through Advertising
Following a single advertising dollar through the modern media ecosystem reveals how many parties depend on it - and how differently the economics work depending on which medium is being used.
$1,000 in direct TV network buy: The network receives approximately $800-900 after agency commission (typically 15% to the media agency). There are no technology intermediaries, no fraud concerns, and no verification complexity. The advertiser received perhaps 15,000-30,000 impressions in a premium, guaranteed environment - at a far higher CPM but with much greater confidence in the quality of those impressions.
How Google makes $200+ billion from advertising
Google's advertising business is one of the most profitable enterprises in human commercial history. Its dominance stems from controlling multiple layers of the ad stack simultaneously: it operates the world's dominant search engine (Google Search), the dominant video platform (YouTube), the dominant ad serving technology (Google Ad Manager), the dominant DSP (DV360), the dominant mobile operating system (Android), and the dominant browser (Chrome) - giving it unparalleled data and inventory reach. Google takes a cut as publisher (when ads appear on its own properties), as technology vendor (when publishers and advertisers use its platforms), and as the auctioneer in the exchange where it effectively represents both buyers and sellers.
The EU's Digital Markets Act and multiple US antitrust investigations have focused precisely on this vertical integration - the argument being that Google's simultaneous presence on the buy side, sell side, and exchange layer creates conflicts of interest that disadvantage publishers and advertisers who use its platforms.
How publishers monetize their audiences
Publishers - whether a television network, a newspaper, a website, or a podcast - are essentially in the business of aggregating audiences and selling access to them. Their revenue per audience member varies enormously by the quality of that audience, the measurement infrastructure available, and the competitive landscape. A premium business publication reaching CFOs can charge $50+ CPM because each reader is a coveted, difficult-to-reach decision-maker. A general news site must compete with millions of other sites for programmatic dollars, typically receiving $2-5 CPM.
The ad-revenue model for publishers has been significantly compressed by the concentration of digital advertising dollars in Google and Meta. A decade ago, the open web's programmatic marketplace was seen as a democratizing force for publisher monetization. Instead, Google and Meta absorbed approximately 50-55% of all US digital ad spending, leaving the rest of the publishing ecosystem competing for the remainder. This dynamic has driven the shift toward subscription and membership models among premium publishers (The New York Times, The Atlantic, The Guardian) who have concluded that ad revenue from the open web cannot sustainably fund quality journalism.
14 - FutureWhere Advertising Is Heading
The cookieless future and first-party data
The deprecation of third-party cookies - already complete in Firefox and Safari, and progressing in Chrome - is the single largest structural change to digital advertising in a decade. The targeting and attribution capabilities built on cross-site cookie tracking must be rebuilt on first-party data (information brands collect directly from their own customers through loyalty programs, email subscriptions, and on-site behavior), contextual targeting (matching ad content to page content rather than user history), and privacy-preserving measurement techniques like clean rooms (where advertisers and publishers can match their data without exposing individual-level records) and media mix modeling (statistical modeling of how advertising drives business outcomes at an aggregate level).
Retail media: the fastest growing channel
Retail media networks - advertising platforms operated by retailers that allow brands to advertise within the retailer's own digital properties using the retailer's first-party purchase data - are the most significant new advertising channel to emerge in the past decade. Amazon Advertising, which generates over $50 billion annually, is the dominant example. Walmart Connect, Kroger Precision Marketing, Target's Roundel, Home Depot's Orange Apron, and dozens of other retailers have built advertising businesses that collectively represent a $120+ billion global market in 2024.
The appeal to advertisers is powerful: retail media lets brands advertise to consumers at the moment of purchase intent (searching for products on Amazon), with targeting based on actual purchase history rather than behavioral inference, and attribution that is closed-loop and verifiable (did the person who saw the ad actually buy the product?). The appeal to retailers is equally clear: advertising margins are dramatically higher than retail product margins, and the business requires no inventory investment.
AI-generated and personalized creative
The production of advertising creative - historically one of the most labor-intensive and expensive parts of advertising - is being transformed by generative AI. Systems that can generate campaign imagery, write ad copy variations, produce video content, and dynamically assemble personalized ads from component pieces are enabling advertisers to produce thousands of creative variations at minimal cost. The implications for testing (creative A/B testing becomes dramatically cheaper) and personalization (ads can be adapted in real time to local context, weather, language, and individual user characteristics) are significant. The implications for employment in advertising agencies and creative production houses are more uncomfortable.
Attention measurement: beyond impressions
The advertising industry is progressively moving from measuring opportunity to see (an ad was in-view on a page) to measuring actual attention - using eye-tracking data (from webcam-based research panels), scroll behavior, and time-in-view signals to establish how much genuine human attention was directed at an ad. Companies like Lumen Research and Adelaide are building attention-based metrics that better predict advertising effectiveness than raw impression counts. If attention measurement becomes standardized, it could fundamentally reprice advertising inventory - rewarding high-attention environments (quality journalism, linear TV viewed without distraction) over low-attention environments (cluttered programmatic display pages) regardless of raw audience size.
The consolidation of streaming and its advertising implications
As the streaming wars produce consolidation (mergers, platform closures, bundling) and all major streaming platforms move to ad-supported models, streaming advertising is rapidly becoming the functional successor to linear TV. By 2027, most projections show CTV/streaming surpassing linear TV in total video advertising revenue in the US. The resulting landscape will look structurally similar to the current programmatic display market - multiple platforms, programmatic buying through DSPs, audience-based targeting - but with the premium content environment and full-screen engagement of television. Advertisers who learn to navigate CTV today are positioning themselves for the dominant video advertising medium of the next decade.
Despite all the technological transformation - from Upfronts to programmatic auctions, from Nielsen panels to attention measurement, from mass market to hyper-targeted individual - the fundamental value exchange in advertising has not changed in 200 years. Someone has content that attracts an audience. Someone else wants to communicate with that audience. The media entity sells the connection between the two. Every innovation in advertising is, at its core, a refinement of that exchange: reaching more people, reaching the right people, reaching them more efficiently, and knowing more precisely whether it worked.
05 - Social MediaMeta, TikTok, YouTube & the Attention Economy
Meta (Facebook & Instagram) - the data-targeting giant
Meta's advertising platform is built on what is arguably the richest behavioral and demographic dataset ever assembled for commercial purposes. Because users voluntarily share their age, location, employer, relationship status, interests, and social connections - and because Meta tracks their behavior on-platform and, through its pixel technology, on millions of third-party websites - advertisers can target with extraordinary specificity: "Women aged 28-35, living within 25 miles of Chicago, who are engaged, have expressed interest in wedding planning, and visited a bridal website in the past 30 days."
Meta generates approximately $130 billion in annual ad revenue, primarily from these targeting capabilities. Campaigns can be optimized for a spectrum of objectives - from reach and brand awareness, through traffic and lead generation, to direct purchases and app installs. The platform charges through a combination of CPM (impression-based) and oCPM (optimized CPM, where the algorithm shows ads to users most likely to take the desired action). Meta's Lookalike Audiences feature allows advertisers to upload their customer lists and have Meta find users with similar behavioral profiles - one of the most powerful prospecting tools in digital marketing.
The watershed moment for Meta's advertising came with Apple's App Tracking Transparency (ATT) policy in 2021, which required apps to ask for permission before tracking users across apps and websites. A significant majority of iOS users declined tracking, cutting off a large portion of Meta's signal for conversion attribution and audience targeting. Meta estimated the impact at $10 billion in lost 2022 revenue, and the company has since invested heavily in privacy-preserving measurement approaches and AI-driven optimization that can perform without granular individual-level tracking data.
TikTok - the engagement engine
TikTok represents a fundamentally different advertising model from Meta. Where Meta's targeting is based on declared interests and social graph connections, TikTok's algorithm is purely behavioral: it learns what keeps each individual user watching, and serves content (and ads) accordingly. This creates an environment where advertisers can reach audiences based on behavioral engagement patterns rather than demographic or declared interest targeting.
TikTok's ad formats are heavily native - In-Feed Ads appear in the For You feed and must compete with organic content for attention, incentivizing high-quality creative production. TopView (a full-screen takeover shown to users when they open the app) commands significant CPMs ($50-$100+) but guarantees unmissable placement. Branded Effects and Spark Ads (boosting organic creator content as paid ads) are unique to TikTok's creator-centric ecosystem. The platform's rapid growth among Gen Z makes it increasingly indispensable for brands targeting young consumers - though its regulatory status in Western markets remains uncertain.
YouTube - video at scale
YouTube, owned by Google, is the world's largest video advertising platform. Its skippable in-stream ads (the 5-second-then-skip format) charge advertisers only when a user watches at least 30 seconds or the full ad - meaning a user who skips at second 6 costs the advertiser nothing. This performance-friendly model makes YouTube attractive for reach with cost control. Non-skippable in-stream ads (maximum 15 seconds) guarantee full exposure. Bumper ads (6 seconds, non-skippable) are used for frequency and reminder messaging in multi-touch campaigns. YouTube's targeting leverages Google's search and browsing data, making it possible to serve video ads to people based on what they've searched for, not just what they've watched.