Advertising
in a Hurry

The Whole Thing in One Page

Advertising is persuasion with a distribution budget. A seller, cause or institution pays to place a message in front of people who did not ask for it, or to appear at the moment they did. The visible output is the poster, film, search result or sponsored post. The industry underneath has to solve three harder problems: who gets the message, what survives in the mind, and whether any profitable change happened because the message existed.

That makes advertising less like hypnotism than probability engineering. An impression can be delivered without being noticed. An ad can be noticed without being remembered. It can be remembered without changing preference, and it can be credited for a purchase that would have happened anyway. Yet small shifts matter when they are repeated across millions of encounters. The business was built on turning those uncertain shifts into something that could be planned, bought and sold.

Its history follows the media that made audiences tradable. Newspapers sold columns and classified notices. Industrial brands needed strangers to recognise packaged goods beyond the shopkeeper who once vouched for them. Agencies began as brokers of space and became manufacturers of strategy, copy, images and research. Coupons and mail order made response countable. Radio and television let sponsors buy national fame. The creative revolution showed that a commercial message could entertain, surprise and become part of culture. Databases divided mass audiences into lists. Search sold access to declared intent. Social platforms inferred likely interests from behaviour. Programmatic systems turned a possible glance into an automated auction.

The advertiser still has two basic jobs. It must make the right people more likely to encounter the message, and it must leave something useful behind. Sometimes that residue is a reason to act now. Often it is lighter: a familiar name, a distinctive shape, a remembered claim, a feeling of fit, a category cue. Advertising changes minds most reliably by changing what comes to mind.

This is also why targeting is less magical than it sounds. Audiences are assembled from geography, context, past behaviour, customer records and statistical prediction. Those signals can improve efficiency, but they can also select people who were already likely to buy. The more perfectly a campaign finds existing demand, the easier it is to confuse prediction with persuasion.

Measurement therefore needs a counterfactual. Attribution asks which visible touchpoint receives credit. Causal measurement asks what would have happened without the advertising. The second question is harder, because the missing world cannot be observed directly. Randomised holdouts, experiments and careful econometrics approximate it, often with uncomfortable uncertainty.

The deepest consequence arrives when advertising pays for the medium itself. Newspapers, broadcasting, search, social networks and much of the open web learned to organise themselves around attracting, classifying and monetising attention. The industry built to change your mind ended up helping to build the places in which minds now spend their time. That is why advertising cannot be understood only by studying famous campaigns. The message, the market that distributes it, the evidence used to judge it and the medium financed by it are one connected system.

That is the book.

Why You Should Care

Open a search engine and type the name of an expensive product. Before the page appears, software may run an auction among advertisers who want access to that moment. Open a social feed and another allocation system is deciding whether an ad should occupy the next slot, which advertiser should get it, how likely you are to respond and what that opportunity is worth. You see a labelled result. Behind it is a market making a prediction about you.

Advertising matters because this machinery sits inside ordinary life. It funds journalism, entertainment, creators, free software, maps, search and social networks. Alphabet's 2025 annual report showed that advertising still supplied well over two thirds of its revenue. Meta's 2025 filing showed an even tighter dependence: advertising accounted for almost all revenue. Those figures do not prove that every product decision is made for advertisers. They do establish the economic pressure under which the products operate. A service funded by subscriptions has to keep subscribers. A service funded by advertising has to keep users, attract commercially useful attention and persuade advertisers that access to it is worth buying.

That bargain has benefits. Advertising subsidises media that many people would not buy directly. It allows a small company to reach buyers without owning a newspaper, shop network or sales force. Search advertising can connect a person who has declared a need with a supplier within seconds. Mass advertising can make a new product known quickly enough for distribution and production to scale. Public information campaigns use the same tools to promote vaccination, road safety or recruitment.

The bargain also has costs. The money follows attention unevenly. Media that attract desirable audiences can command high prices even when the content is cheap to produce; difficult public-interest reporting can struggle even when it is socially valuable. Platforms gain an incentive to collect signals that make targeting and measurement more valuable. Publishers can be tempted towards volume, outrage or clutter. Influencers blur entertainment and promotion. A market that rewards measurable response may undervalue effects that are delayed, diffuse or hard to attribute.

You should also care because the popular defences against advertising are poor. One defence is fear: the idea that skilled advertisers possess a secret route into the unconscious and can make people want anything. That gives the industry too much credit. Products fail despite enormous budgets, famous campaigns often cannot prove their incremental return, and people ignore most commercial messages. The opposite defence is contempt: the belief that advertising cannot work on anyone intelligent enough to recognise it. That gives memory too little credit. You can dislike an ad and still remember the name. You can reject the argument and still become more familiar with the brand. You can believe you chose freely while choosing from a set of options whose availability in memory was partly purchased.

For anyone who spends money on advertising, the subject becomes practical. What are you buying: reach, attention, clicks, leads, sales, fame, reassurance? Which of those can the medium deliver? Which can it merely report? What would the customer have done without the campaign? A dashboard can answer the wrong question with six decimal places. A good operator learns to distinguish delivery from attention, attention from memory, memory from action, and action from incremental profit.

For everyone else, advertising is useful training in institutional scepticism. An advertiser wants profitable change. An agency may be paid for labour, media, performance or retained expertise. A publisher wants inventory sold. A platform wants auction demand and continued use. A measurement vendor wants measurable events. The person seeing the ad wants relevance, entertainment, a bargain or escape. No villain is required for bad incentives to produce clutter, invasive tracking, weak claims or flattering metrics.

Advertising is also a cultural archive. Old campaigns expose assumptions about status, gender, family, work, health and normality because persuasion begins with a theory of what the audience values. Ads do not single-handedly create those values, but repeated commercial images can reinforce, challenge and reveal them.

The subject is larger than commercials because advertising changes the environments around them. Once attention can be sold, someone has reason to manufacture more of it, package it more precisely and prove that it caused something. That logic reaches from the newspaper page to the television schedule to the infinite feed. Learn the logic and the modern media world becomes easier to read.

The Core Ideas

1. Advertising buys distribution before it buys persuasion

The first advertising problem is not what to say. It is how to get the message into the path of someone who might matter.

That sounds obvious until the language of digital advertising hides it. An advertiser may buy impressions, clicks, completed video views or conversions, but the basic media transaction still begins with access to an opportunity. A newspaper sold space. A broadcaster sold time around programmes. A search engine sells eligibility to appear against a query. A social platform sells a chance to occupy a slot in a feed. The object being traded changes, but the advertiser is buying distribution of a message.

Distribution is valuable because attention is scarce and uneven. A million people can pass a billboard without giving it a conscious thought. A small trade publication can matter more than a mass audience if its readers control expensive purchasing decisions. A person searching for "emergency plumber" may be worth far more to a local plumber than the same person watching sport. The commercial value of attention depends on who holds it, why they are there and what decisions are near.

This creates the first distinction a newcomer needs. Delivery is not attention. In digital systems an impression normally means that an ad was served according to the platform's counting rules. Viewability adds a technical condition about whether enough of the ad had a chance to appear on screen. Attention is a human event. Memory and behaviour come later. Treating those stages as interchangeable is how advertisers pay for numbers that look like outcomes.

Media planning is the discipline of arranging enough useful opportunities at an acceptable price. Reach asks how many different people received a chance to encounter the message. Frequency asks how often. Context changes what those encounters mean. Timing can make a weak-looking placement valuable because it arrives near a decision. Geography can matter more than demographic precision. The planner's task is to combine these dimensions without mistaking a cheap impression for a valuable one.

Mass media changed advertising because they made distribution scalable. A manufacturer that once depended on local merchants could buy national recognition. The manufacturer did not have to persuade every person deeply. It could create enough familiarity that a packet, soap, camera or car arrived in the shop with demand already attached to its name. Advertising became a way to reduce the distance between anonymous producer and anonymous buyer.

Digital media changed the unit again. Instead of buying a magazine issue or television audience as a block, advertisers could increasingly buy individual opportunities selected by software. That sounds like the end of waste. It is better understood as a change in where waste hides. A platform can identify a person who looks likely to buy and still fail to change that person's behaviour. It can deliver an ad to the right profile at the wrong moment. It can optimise towards an easy proxy that has little economic value.

The media buy therefore sets a hard ceiling on everything that follows. Brilliant creative cannot influence someone who never encounters it. But abundant exposure to weak or misdirected creative merely wastes money faster. Advertising begins as distribution because persuasion needs a path.

2. The mind is changed mainly through memory, not argument

The stereotype of advertising is a miniature debate. The advertiser makes a claim, the consumer evaluates it and a purchase follows. Some advertising works that way. Much of it does not have enough time, attention or novelty for such a sequence.

Most consumer choices are made under conditions of incomplete attention. A supermarket shelf contains too many products to analyse from scratch. A phone contract renewal arrives while other things are happening. A restaurant decision is made among a handful of names people can retrieve. Advertising can affect these moments without winning an explicit argument. It can make a brand easier to notice, easier to recall, more familiar, more associated with a need and less risky to choose.

That is why memory is central. An ad has to encode something, connect it to the right source and make it retrievable later. The "something" may be a product fact, but it may also be a colour, shape, sound, phrase, character, demonstration or social meaning. The difficult part is not merely being memorable. A joke remembered without the advertiser is entertainment donated by the advertiser. Distinctive assets work because they help the memory point back to the source.

Familiarity has its own effect. Robert Zajonc's classic work on mere exposure showed that repeated exposure can increase liking under some conditions even without elaborate conscious evaluation. Advertising research since then has complicated the picture, but the practical lesson survives: recognition and ease can matter. Familiar options often feel safer, and familiarity can increase the chance that a brand enters the set considered at all.

This does not mean repetition is automatically good. Repetition can become irritating, especially when the same execution follows a person across sites. It can also be redundant. The useful question is whether another exposure adds probability of retrieval or merely adds cost. Frequency should be treated as a dose, not a superstition.

Advertising also attaches meanings to products that their physical features cannot contain. A trainer can be presented as performance equipment, street status or evidence of taste. A car can be framed around safety, rebellion, family competence or technological modernity. These meanings are not imaginary because products are social objects. People use brands to make inferences about themselves and others. The advertiser is competing to shape those inferences.

This is where advertising approaches branding without taking over branding's job. Branding owns the long-term system of identity, reputation and expectations. Advertising rents moments in which parts of that system can be reinforced, refreshed or altered. The campaign is temporary. The memory structure it tries to build may last much longer.

The subtitle promises an industry built to change your mind. The change is often smaller than a conversion and more durable than a click: what you recognise, what you expect, what feels normal, what enters the shortlist, what social meaning attaches to a name. Advertising becomes powerful by accumulating these minor changes across time.

The category itself can be one of those changes. Before instant coffee, deodorant, private pensions or streaming subscriptions become ordinary purchases, people have to learn what kind of problem the category claims to solve and what a normal use looks like. Advertising can therefore educate and normalise at the same time. Early campaigns often explain a new object more explicitly because the buyer lacks a script for it. Mature categories can spend fewer words explaining the object and more effort making one name easier to retrieve than another. The work changes because the buyer's memory changes.

3. Creative strategy is the art of deciding what must survive

A media plan can create opportunities. Creative work determines what those opportunities contain.

This is why "make the ad attractive" is an inadequate brief. The advertiser has a commercial problem: perhaps too few people know the product, perhaps they know it but associate it with the wrong use, perhaps they distrust the price, perhaps a competitor owns the category cue, perhaps existing buyers need a reason to return. Creative strategy turns that problem into a communication task and then into an execution that can survive contact with a distracted person.

The classic agency brief exists to force choices. Who is the audience? What behaviour or perception needs to change? What is the proposition? What evidence supports it? What tone is appropriate? What must the person remember? Different agencies use different forms, and good advertising often violates bad templates, but the discipline is real. A commercial message cannot carry the entire product strategy. Compression is unavoidable.

Claude Hopkins, writing in 1923, treated advertising as salesmanship multiplied by media. His world prized specificity, demonstrations, samples, coupons and measurable response. Later advertising figures pushed against the idea that a rational claim was enough. Bill Bernbach's work at Doyle Dane Bernbach made wit, art direction and respect for the audience central to campaigns such as Volkswagen's 1959 "Think Small". Rosser Reeves argued for a distinctive selling proposition. David Ogilvy combined research, long copy, image and brand character. None discovered the one true formula. Their disagreement is useful because it exposes the jobs creative can perform.

A demonstration reduces uncertainty. A vivid claim supplies a reason. Humour buys attention and social transmission. Emotion can attach a feeling to the source. A distinctive visual system improves recognition. A story can make a product's role concrete. A repeated slogan can compress an association. An odd execution can break through the category's visual sameness. Creativity matters when it improves the probability that the intended mental residue is encoded and correctly attributed.

This is why originality alone is a poor standard. An ad can be original and commercially useless. It can win attention for a celebrity, joke or cinematic scene while the brand disappears. Conversely, dull advertising can be efficient when the task is narrow and intent is already high. A search ad for a locksmith does not need to become cultural history. It needs to be relevant, credible and available at the moment of need.

Creative also operates inside culture rather than outside it. Advertisers borrow symbols, music, humour, aspiration, class cues, gender codes and anxieties that already exist. Successful campaigns can then feed those materials back into culture. Catchphrases escape the commercial. Product categories acquire rituals. Christmas advertising becomes a seasonal media event. Sponsorship ties brands to sport, music and identity. Advertising reflects society, exaggerates parts of it and sometimes helps normalise what it repeatedly depicts.

That social power is harder to measure than a sale, which is one reason it is easy for performance systems to neglect. A campaign may fail to produce an immediate response while changing category language or expectations. Another may deliver cheap clicks while degrading trust. Creative strategy should therefore begin with what must survive in the mind, not with what the platform happens to count.

Production constraints shape the answer. A six-second video, a search headline, a poster passed at speed and a two-minute cinema film are not the same canvas. The message must fit the amount and kind of attention available. This is why transferring a television script into a banner or shrinking a print ad into a social tile so often produces weak work. Media strategy and creative strategy meet at the encounter itself: what can this format realistically make a person notice, understand and retain?

4. Audiences are constructed from signals

Advertisers talk about "the audience" as if it were a crowd waiting to be discovered. In practice audiences are definitions imposed on messy populations.

A local newspaper once offered geography and readership. A women's magazine offered an inferred demographic and set of interests. Direct mail added lists built from addresses, subscriptions and purchases. Television ratings estimated who watched programmes. Digital advertising added enormous quantities of behavioural and transactional signals: searches, page visits, app activity, device data, customer records, content consumed and patterns inferred from them.

None of these signals is the person. They are proxies for something the advertiser cares about. A search for running shoes may indicate purchase intent, curiosity, research for someone else or an existing customer's attempt to find the same model again. A platform category such as "likely to move" or "interested in luxury travel" is a prediction. Even first-party customer data describes what someone did before, not what they will do next.

Targeting is useful when the proxy meaningfully raises the chance that an impression is valuable. Geography can eliminate impossible buyers. Business-to-business advertising can focus on industries or job functions. Retargeting can remind someone who abandoned a basket. Customer lists can suppress existing buyers from an acquisition campaign or help find similar prospects. Search can match supply to explicit demand.

But targeting introduces a measurement trap. Suppose people who visited a product page are far more likely to buy than people who did not. Advertising to those visitors may produce an impressive conversion rate even if the ad changes almost nothing. The targeting system has found probability rather than created it. The campaign looks effective because the audience was already different.

This matters because the digital industry has often sold precision as persuasion. Better prediction can make media buying more efficient, and it can increase reported return by concentrating spend where conversions are easy to observe. Those are not the same achievement. A perfectly targeted ad shown only to people who would buy anyway has excellent prediction and zero incrementality.

There is also a social boundary. Some targeting criteria are intrusive, discriminatory or legally restricted. The European Union's Digital Services Act prohibits online platforms from targeting ads using sensitive categories of personal data and restricts targeting of minors. UK privacy rules governing storage and access technologies impose consent requirements on most commercially viable forms of behavioural online advertising. These constraints are not external annoyances attached to an otherwise neutral machine. They define which signals society is willing to turn into commercial opportunities.

The practical mental model is therefore simple: every audience is a hypothesis. Ask what signals created it, what those signals proxy, who is excluded and whether the group is more persuadable or merely more predictable.

Segmentation also creates political and commercial power because it changes who is visible to whom. A mass television campaign is public: competitors, journalists and regulators can see much of what the population sees. Highly targeted digital campaigns can show different messages to different groups, sometimes with little shared public record. Ad libraries and transparency rules partly respond to that asymmetry. The more individually addressable advertising becomes, the less any single observer can see the whole campaign.

5. Media is a market for moments, with incentives attached

Advertising media looks like a set of channels. Economically, it is a chain of markets.

At one end is the advertiser, with a budget and a desired outcome. At the other is an opportunity for a message to appear before a person. Between them may sit an agency, media owner, ad exchange, supply-side platform, demand-side platform, data provider, measurement company, affiliate network, creator or reseller. Each participant performs a function and has a way of being paid. Those payment rules shape behaviour.

Traditional media buying made the transaction visible. A magazine quoted rates for pages and positions. Television sold spots against programmes and estimated audiences. Outdoor companies sold locations and periods. Buyers negotiated discounts, schedules and packages. The uncertainty concerned who encountered the message and what it achieved.

Digital media added auction logic. In search advertising, eligible advertisers compete for positions against a query, with ranking influenced by bids and measures of expected usefulness or quality. In programmatic display, a page or app can create a bid request, systems can decide whether the impression matches campaign rules, buyers can submit prices and a winner can be selected in fractions of a second. The exact architecture varies by platform and transaction type, but the principle is that software allocates many individual opportunities dynamically.

Auctioning creates efficiency and opacity at the same time. The advertiser can bid differently by context and predicted value. The publisher can sell inventory to a larger pool of buyers. Yet a long chain makes it harder to know how much of the advertiser's pound reaches the publisher, how data is used, which fees are taken and whether the inventory is genuine. The UK Competition and Markets Authority's study of online platforms and digital advertising documented concerns about market power, conflicts and opacity in major parts of this ecosystem.

The incentive problem is broader than programmatic advertising. An agency paid as a percentage of media spend can benefit when spending rises. An affiliate paid per attributed sale may compete aggressively for the last click near purchase. A platform optimised to conversions may favour users most likely to convert whether or not the ad causes the conversion. A publisher paid per impression benefits from producing more inventory. A creator paid for sponsorship must preserve audience trust while satisfying the advertiser.

This does not make the industry fraudulent. It makes contract design part of advertising. The metric used for payment becomes a target. Cheap clicks invite clickbait. Cheap impressions invite low-quality inventory. Cost per acquisition can reward harvesting demand that another activity created. Brand-awareness targets can encourage surveys that are easy to move but commercially weak.

Good media buying therefore means more than finding low prices. It means deciding which moments deserve a price, how much uncertainty is tolerable and whether every participant is rewarded for something close enough to the advertiser's real objective.

There is a second price hidden in media: opportunity cost. The cheapest audience is not automatically the best use of budget. If a platform can deliver a thousand impressions for £3 but those impressions contain little chance of memory or action, a £15 placement in a more valuable context may be cheaper in economic terms. Media metrics turn human opportunities into comparable units, but comparison only works when the units remain connected to the job. Cheap supply can become expensive waste.

6. Measurement begins with the world that did not happen

John Wanamaker is often credited with the line that half the money spent on advertising is wasted but he does not know which half. The attribution of the quotation is uncertain. The problem it describes is not.

A campaign runs and sales rise. Did the advertising work? Perhaps. Perhaps demand was already rising. Perhaps the campaign coincided with a promotion, better distribution, good weather, a competitor's stock problem or seasonal demand. Perhaps the ad reached people who were already going to buy. Observing what happened after advertising is not enough to know what happened because of it.

The missing object is the counterfactual: what would the same market have done without the campaign? Because the same person cannot simultaneously be exposed and unexposed in the same moment, causal measurement has to construct a comparison.

Randomised experiments do this most cleanly. Eligible people, regions or units are randomly assigned to receive or not receive advertising, so pre-existing differences should balance on average. The difference in outcomes estimates incremental effect. Large platforms have made such experiments possible at extraordinary scale. The result has often been humbling. Work by Brett Gordon and colleagues comparing observational methods with Facebook field experiments found that common observational approaches could diverge substantially from experimental estimates. Randall Lewis and Justin Rao showed why precise return-on-advertising estimates can require huge samples when advertising effects are small relative to natural variation in purchasing.

Paid search provides a famous example of selection. Thomas Blake, Chris Nosko and Steven Tadelis ran large experiments at eBay and found that much branded search advertising was capturing users who would have reached eBay anyway. That result does not prove search advertising is generally useless. It proves that a channel can appear powerful because it sits close to intent.

Attribution solves a different problem. If a customer saw a video, clicked a search ad and later bought, a reporting system may assign credit to the last click, divide credit across contacts or use a model to distribute it. This can help describe paths and allocate internal credit. It does not by itself establish causation. The rule decides who gets credit among observed touches; it does not reveal whether the sale needed any of them.

Measurement also has time horizons. A campaign can create an immediate sale, a later memory, a price expectation, future search, retailer demand or an effect on existing customers. Short experiments can miss long effects. Long studies face more confounding. Brand surveys can detect movement in awareness or consideration but may not map cleanly to profit. Econometric models can use historical variation but depend on assumptions. There is no universal instrument that turns advertising into a certain return figure.

The correct response is not measurement nihilism. It is hierarchy. Count delivery to manage delivery. Use diagnostic metrics to improve execution. Use experiments where feasible to estimate incrementality. Tie results to contribution margin and long-term economics rather than revenue alone. Most important, keep the question intact: compared with what would otherwise have happened?

This is why "return on advertising spend" deserves caution. Revenue divided by spend ignores product margin, fulfilment costs, repeat purchasing, cannibalisation and the possibility that some attributed customers were not incremental. A campaign can report £5 of revenue for every £1 spent and still destroy profit. Another can look weak in a thirty-day attribution window while creating customers whose value arrives later. The economic question is the incremental profit created over an appropriate horizon, net of the costs required to obtain it.

One more complication matters in mature advertising systems: effects can spill across channels. A television campaign may increase branded search; an outdoor campaign may make a later sales call easier; a creator partnership may improve conversion on a retailer site where the final sale is recorded. Channel-by-channel attribution tends to divide a system that consumers experience as one sequence of encounters. Experiments and modelling should therefore be designed around business decisions, not around preserving neat ownership of credit between departments.

7. Advertising became part of the architecture of media

The first Core Idea described advertising as paid distribution. Follow that transaction far enough and it begins to redesign the distributor.

A newspaper with advertising revenue can charge readers less because advertisers also pay for access to them. Commercial radio and television turned programmes into audience-generating assets surrounded by sponsor messages. Search engines offered extraordinarily useful services without charging most users per query because commercial intent could be auctioned. Social networks could scale globally while remaining free at the point of use because user attention and behavioural signals supported advertising markets.

Once the medium depends on advertising, it has an incentive to produce more sellable moments. That can mean better content, larger audiences and useful free services. It can also mean more pages, more breaks, more notifications, more recommendation loops and stronger data collection. The funding model does not dictate every design choice, but it changes what counts as economically valuable.

Digital platforms deepened the loop because the medium, the audience database, the auction and much of the measurement system can sit inside the same company. A platform observes behaviour, predicts which ad may be valuable, allocates the impression, records the response and reports performance back to the advertiser. This integration is efficient. It also means the seller may control much of the evidence used to evaluate the sale.

The social consequence is often described as "the attention economy", but attention alone is too narrow. What platforms monetise is a stream of classified opportunities: this person, in this context, with these inferred interests, available now. More observation can improve those classifications. More engagement can create more opportunities. Better measurement can reassure advertisers and increase demand. The commercial loop links product design, data collection and advertising economics.

Regulation has responded by drawing lines around claims, identification and data. UK advertising codes require marketing communications to be recognisable as such and prohibit misleading advertising under detailed rules. The Digital Markets, Competition and Consumers Act 2024 strengthened the UK's consumer-protection enforcement framework for unfair commercial practices from April 2025. The Information Commissioner's Office's 2026 guidance states that storage and access technologies used for online advertising require consent under the current framework. In the European Union, the Digital Services Act requires greater ad transparency and restricts targeting based on sensitive data and minors' personal data.

Those rules can change practices without changing the business logic. If a medium is financed by advertising, someone must continue persuading advertisers that access to its users is worth money. The architecture adapts.

This completes the loop. Advertising began by renting space inside media. As the revenue became important, media organisations learned to create, package and optimise the spaces being rented. The industry built to change minds therefore changed the institutions that compete for minds in the first place.

The same logic explains why debates about advertising quickly become debates about media power. The question is no longer confined to whether a shampoo commercial is persuasive. It includes who can buy visibility, which behaviour becomes legible to the seller, which content receives subsidy and which commercial signals influence recommendation systems. Advertising remains a business function, but at sufficient scale its infrastructure becomes part of public culture.

How It Actually Works

Before there was an industry

People have advertised for as long as sellers have needed strangers to notice them. Ancient notices, shop signs, criers and handbills all performed recognisable advertising jobs. They announced availability, differentiated sellers and directed people towards a transaction. What they lacked was the modern industrial combination: mass production, mass media, professional intermediaries, branded goods and systematic measurement.

Printing expanded the amount of commercial space that could be sold. Newspapers and periodicals placed sellers beside news and entertainment, creating the basic two-sided bargain that still runs through much media. The publisher gathered readers; advertisers paid for access to them. Classified notices made the exchange explicit. Display advertising turned the page itself into a designed commercial surface.

The agency starts as a broker

The modern agency did not begin as a temple of creativity. It began by arranging media.

In nineteenth-century America, agents bought or represented newspaper space and resold access to advertisers. Duke University's advertising-history collections trace firms such as the agency that became J. Walter Thompson from this brokerage world. N. W. Ayer & Son, founded in Philadelphia in 1869, became influential as agencies added planning, copy and broader client service. The intermediary gradually shifted from selling the medium's inventory towards representing the advertiser's communication problem.

That shift mattered because the work became specialised. Someone had to choose media, write copy, commission illustrations, negotiate rates, research markets and keep campaigns coherent across publications. The agency bundled skills that a manufacturer might not maintain internally. It also introduced the principal-agent problem that remains today: the adviser can possess more information about media than the client while being paid through arrangements connected to media spend.

Packaged goods create the need for strangers to trust names

Industrial production changed what advertising had to accomplish. A local merchant could once recommend loose goods whose producer was invisible to the buyer. Packaged goods moved the manufacturer's name to the front. The producer now wanted a shopper to ask for a specific soap, biscuit, camera or patent medicine before the shopkeeper could redirect the sale.

Brands therefore became commercial shortcuts. Packaging made a product recognisable. Advertising made the recognition travel. National magazines, newspapers, posters and catalogues could expose people to the same names repeatedly across large territories. The manufacturer gained some power over demand instead of depending entirely on retailers.

That power altered competition. A heavily advertised brand could arrive at a wholesaler or retailer with consumers already asking for it. The retailer still controlled shelf space, but the producer had created a constituency outside the shop. Advertising thus helped manufacturers build assets that were not factories or patents: recognisable names with demand attached. That commercial development belongs fully to branding in the long run, but advertising was one of the main engines that carried those names across distance.

The period also produced excess. Patent medicines made extravagant health claims. Testimonials, fear and pseudo-science flourished. Advertising's ability to scale a truthful claim also scaled deception. Modern regulation grew partly from the need to make commercial speech legible and accountable when the seller and buyer no longer knew each other.

Direct response tries to make the invisible visible

Mail order introduced a different ambition: trace the response.

A newspaper ad could include a coupon, code or address. Different versions could run in different publications. Orders could be counted. Claude Hopkins built his reputation in this environment and later codified its logic in Scientific Advertising. Test the headline. Offer a sample. Use specific claims. Give the reader a reason to act. Track which advertisement produced replies.

This was a major advance, but the limitation should be clear. Response tracking is not the same as causal measurement. A code shows that a customer used the code. It does not prove the customer would otherwise have done nothing. The distinction between attribution and incrementality existed before digital dashboards; digital systems merely made the attributed path more detailed.

Mail order also made copywriting unusually accountable. A vague institutional message could survive because nobody knew which sentence mattered. A direct-mail package that failed to produce orders was harder to romanticise. That pressure produced a tradition obsessed with offers, headlines, guarantees, specificity and testing. Modern performance advertising inherits both its strength and its bias: it is excellent at optimising things that produce visible response and can neglect effects whose value appears later or elsewhere.

Radio sells time and sponsorship

Broadcasting changed the unit from space to time. In the United States, commercial radio developed in the early 1920s, with WEAF in New York commonly associated with the sale of airtime in 1922. Networks then allowed programmes and sponsor messages to reach large audiences simultaneously.

Early broadcasting often tied sponsors closely to programmes. Advertising agencies could be involved in producing entertainment on behalf of clients. The advertiser was therefore funding more than interruptions. It could help finance the programme itself. The association survives in language such as "soap opera", reflecting the heavy sponsorship of daytime serials by soap and household-product companies.

Radio made voice, music, repetition and personality available to advertisers at national scale. It also changed the social position of the advertisement. A print ad occupied its own rectangle. A sponsor could become part of the programme's identity, with products named in introductions, performers delivering commercial messages and agencies shaping entertainment around client needs. The line between financing culture and entering culture became porous.

It also showed the cultural bargain more clearly than print. Audiences received entertainment without paying the full cost because sponsors wanted the audience assembled around it.

Television makes fame purchasable

Television intensified the arrangement by adding image, movement and domestic reach. In early American television, programmes were often produced with deep sponsor and agency involvement. Over time, networks strengthened control of schedules and the market moved towards multiple spot advertisements within and between programmes.

Audience ratings became a common commercial currency. They did not reveal individual persuasion. They estimated how many people a programme could deliver, allowing buyers and sellers to price access. The system separated the purchase of distribution from the uncertain effect of the creative message.

Television's scarcity mattered. A limited number of channels concentrated attention. National advertisers could repeatedly reach huge shares of the population with the same work. That allowed campaigns to create common cultural references in a way fragmented digital media finds harder. Advertising could manufacture fame because the medium manufactured simultaneous audiences.

The scarcity also concentrated bargaining power. A national advertiser needed the networks; the networks needed major advertisers. A small number of channels made schedules comprehensible and expensive. Today's media buyer faces the opposite problem: almost unlimited inventory, far more fragmented attention and software required to make the market manageable. The industry's operational complexity grew as media abundance replaced media scarcity.

The creative revolution changes the tone

By the late 1950s and 1960s, a reaction formed against formulaic, repetitive and research-heavy advertising. Doyle Dane Bernbach's Volkswagen work became the emblem. "Think Small" placed a tiny car in a field of white space while American car advertising was celebrating size and chrome. The campaign did not abandon selling. It sold through a different social posture: wit, visual restraint and an assumption that the reader could get the joke.

The period matters less because one agency "invented creativity" than because it made a continuing tension visible. Advertising can treat the receiver as a target to be struck with claims, or as an audience whose attention has to be earned. Good work often does both: it is commercially disciplined and culturally alert.

As agencies accumulated creative prestige, advertising also became an industry with its own status system, awards, famous executives and mythology. That system could encourage better craft. It could also reward work that impressed other advertisers more than customers.

Databases fracture the mass audience

Direct mail, loyalty schemes, credit records and customer databases gradually made audiences more addressable. Instead of buying everyone who read a publication, advertisers could buy or build lists defined by past behaviour, geography or demographic characteristics. Catalogue companies and financial marketers became skilled at scoring and selection.

The change was conceptual. Mass advertising asked which medium gathered the right kind of people. Database advertising asked which people should receive the message. The audience moved from being a property of the publication towards being a property of the data.

That movement changed agency work too. Media planning could become less about selecting culturally coherent environments and more about locating defined people wherever inventory was available. The trade-off still exists. Buying the person can improve efficiency; buying the context can supply meaning, attention and trust. An airline ad beside travel reporting is doing something different from the same ad delivered to the same reader inside an unrelated mobile game.

This also changed privacy. A newspaper reader could see an ad without the advertiser knowing who they were. A database campaign depended on records linked to people or households. The commercial value of information became more obvious long before social media.

The web creates the clickable impression

The commercial web added a new object: an ad that could be served, seen, clicked and followed to an immediate destination. HotWired's AT&T banner in October 1994 is conventionally treated as the first web banner ad. It became famous partly because novelty produced a click-through rate that later display advertisers could only dream of.

The click looked like a revolution in accountability. Print had coupons; television had ratings and sales models; the web could record individual interactions automatically. That produced a powerful belief that digital advertising had solved the old measurement problem.

It had solved a logging problem. The system could record more of what happened after exposure. It had not automatically learned what would have happened without exposure.

The web also multiplied inventory. Almost any page could carry advertising, and publishers could create more pages cheaply. Abundance pushed prices down for generic display space and increased the importance of data, targeting, format and measurement as sources of differentiation. Where television sold scarce national attention, the web often sold plentiful fragments of attention. That economic difference helps explain both cheap reach and chronic clutter.

Search sells intent

Search advertising improved the match between commercial messages and immediate need. A query is unusually valuable because the user has volunteered a signal. Someone searching "running shoes size 10" or "accounting software for small business" is easier to match with a relevant advertiser than a random member of a television audience.

Pay-per-click pricing also made the transaction feel cleaner. The advertiser could pay when someone clicked rather than for every exposure. Auctions let competing advertisers bid for scarce positions. Google and other platforms developed ranking systems in which bid price interacts with predicted relevance or usefulness, so the highest bidder does not mechanically win every slot.

Search remains one of advertising's most powerful designs because the advertising is close to the user's task. Its weakness follows from the same fact. People who search for a brand or product may already be highly likely to buy. The closer advertising sits to demand, the easier it is to harvest demand and call the harvest creation.

Social media turns behaviour into an audience model

Social platforms collected a richer stream of signals because users were continuously expressing relationships, interests, location, content preferences and behaviour. Advertisers could define audiences through platform tools, upload customer lists, retarget visitors and ask algorithms to find people similar to existing buyers.

The feed also changed the creative contest. An advertisement no longer competed only with other ads. It competed with friends, news, celebrities, humour, outrage and whatever else the recommendation system believed could hold attention. Advertising adopted the visual grammar of the feed: short video, creator endorsements, native formats and rapid iteration.

The strongest platforms combined audience construction, media inventory, auction, optimisation and measurement. This reduced friction for advertisers. It also concentrated information. The same company could decide who should see the ad, report who did see it and estimate what the ad achieved.

Creator advertising added another layer. A sponsored post can borrow the creator's tone, audience and accumulated trust, making the advertisement feel less like a separate interruption. That can improve relevance and attention, but it makes disclosure important. The audience needs to know when a recommendation is commercially motivated, which is why UK advertising rules devote specific attention to recognisability in influencer and affiliate marketing.

Programmatic systems automate the trade

Outside the largest closed platforms, programmatic advertising connected advertisers and publishers through software. A publisher could expose an impression to an exchange. Buyers could evaluate contextual and audience signals through demand-side systems. Supply-side systems could manage the publisher's inventory. Auctions could happen while the page loaded.

The simplified story is seductive: every impression goes to the buyer who values it most. Real markets contain complications. Inventory quality varies. Intermediaries take fees. Fraud can create fake impressions or fake users. Buyers may not know every site on which their ads appear. Publishers may not know the full path by which their inventory is resold. Privacy rules constrain which signals can be collected and shared.

The technological achievement is still remarkable. Advertising turned an uncertain human glance into a machine-readable object that can be classified, priced and traded in milliseconds.

It also created new classes of quality control. Advertisers worry about brand safety, whether an ad appears beside material that damages the brand; publishers worry about malicious or misleading ads; both worry about invalid traffic and fraud. Automated scale means every error can spread faster. The modern media operation therefore includes verification, exclusion lists, suitability rules and fraud detection alongside the buying itself.

Programmatic markets also changed the skills inside advertising. Media buyers increasingly needed analysts, traders, engineers and privacy specialists beside negotiators and planners. Creative teams began producing many variants because software could test and rotate them. The industry that once organised itself around publications and programmes increasingly organised itself around data flows, platforms and optimisation systems.

That change also weakened the old assumption that the audience and the medium were inseparable. A television viewer could now be targeted through connected television, a supermarket shopper through retail-media networks and a podcast listener through dynamically inserted audio. The same person could become different inventories depending on the commercial data available in each setting. The unit of planning shifted from "which publication reaches our buyer?" towards "which opportunity involving this buyer is worth purchasing?"

The present bargain

The contemporary industry mixes all previous eras. Television still sells reach. Posters still sell location. Direct mail still uses lists. Search sells intent. Social media sells prediction and scale. Retailers sell advertising against their own shopper data. Influencers sell trust and access to communities. Streaming services alternate between subscription and advertising models. Brands use broad campaigns and narrow response campaigns together.

The strongest current shift is not the death of one medium but the convergence of media with data and commerce. Retail media brings advertising close to purchase records. Connected television combines the screen of broadcasting with addressable delivery. Platforms automate creative variants and bidding. Generative systems can produce more advertising material faster, increasing the value of good strategic judgement because production itself becomes cheaper.

The legal bargain is also changing. In the UK, marketers remain responsible for identifiable and non-misleading advertising under the CAP and BCAP systems and consumer law. The ICO's April 2026 guidance states that online advertising uses of storage and access technologies require consent under the current rules, despite newer statutory exceptions for other lower-risk purposes. In the EU, the Digital Services Act adds transparency duties and bans specified forms of targeted advertising involving sensitive data and minors.

The industry keeps adapting because the underlying demand remains: sellers want paid access to buyers, media owners want money for access to audiences, and both want evidence that the exchange changed something.

That continuity is easy to miss because the vocabulary changes quickly. "Native", "retail media", "connected television" and "creator partnerships" can sound like separate revolutions. Most are recombinations of old functions: gather attention, define an audience, insert a persuasive message, price the opportunity and measure response. New technology matters, but the enduring structure is what makes the history useful.

How we know

Advertising leaves an unusually visible archive. Newspapers, magazines, posters, radio recordings, television commercials, agency files, campaign memoranda and corporate records preserve what the industry made and how it talked about its work. Duke University's Hartman Center, the Smithsonian and the Library of Congress hold major collections that make the institutional history unusually concrete.

Effect is harder. Preserved ads show what people were exposed to, not what the exposure caused. Sales records, clicks and customer journeys are observational unless a credible comparison separates advertising from prior demand and other changes. Randomised field experiments provide stronger causal evidence, but even huge experiments can be noisy because the incremental effect of one campaign may be small beside ordinary variation in purchasing. The history of advertising is therefore easier to observe than its average power. Confidence should be highest about the machinery and lower about universal claims of effectiveness.

What People Get Wrong

“Advertising is mind control”

The fantasy persists because it is flattering to both sides. Advertisers get to imagine hidden power. Consumers get to imagine that any influence must have bypassed their rational selves.

The famous subliminal-advertising story is the cleanest example. In 1957 market researcher James Vicary claimed that imperceptible messages such as "Eat Popcorn" and "Drink Coca-Cola" had been flashed during cinema screenings and produced dramatic sales increases. The demonstration became cultural folklore. Vicary later admitted that the supposed experiment had not been conducted as claimed. Later laboratory research has found narrower subliminal priming effects under constrained conditions, including work showing that a subliminal drink brand could influence choice among participants who were already thirsty. That is a long way from installing desires in people who do not have them.

Ordinary advertising works through exposure, memory, expectations, social meaning, information and timing. These mechanisms can influence choice without coercing it. The correction matters because exaggerating advertising's power hides the conditions under which it is weak: bad products, high prices, poor availability, hostile experience and people who are not in the market.

“If people remember the ad, it worked”

Memorability is necessary for some advertising jobs and irrelevant for others. It is never sufficient.

A funny film can be remembered while the advertiser is forgotten. A shocking execution can make the brand famous for the wrong reason. A campaign can lift prompted awareness while failing to increase profitable demand. Conversely, a plain search ad can work commercially without leaving any lasting cultural memory because its job was to provide a credible route at a high-intent moment.

The useful question is what was remembered and whether that memory was attached to the correct source and decision. Creative research often distinguishes attention-getting novelty from brand linkage for this reason. Advertising needs retrievability, not applause in isolation.

The correction matters because the industry has many intermediate metrics that are easy to celebrate. Recall, views, shares and awards can diagnose parts of the communication task. None automatically substitutes for the commercial or behavioural job the campaign was meant to do.

A useful creative test therefore has two axes: did the work leave the intended trace, and did that trace belong to the advertiser? High scores on only one axis are dangerous. Anonymous entertainment wastes brand memory; unmistakable but forgettable work wastes attention.

“Targeting finds the people advertising can persuade”

Targeting finds people who match signals. Persuadability is a different property.

A platform can identify users who visited a product page, searched a category or resemble existing buyers. Those people may be excellent prospects because they already want the product. Showing them an ad can then produce high conversion rates even when the incremental effect is small. This is selection, not failure: the system may be doing exactly what it was asked to do. The error appears when the advertiser treats the selected group's behaviour as proof of persuasion.

Some targeting does create real efficiency by excluding impossible buyers or concentrating on contexts where the message can matter. The point is not that targeting is useless. It is that prediction and causal influence are separate achievements.

The correction matters because the more sophisticated the audience model becomes, the more convincing the dashboard can look. Precision in classification does not guarantee precision in persuasion.

“Clicks tell you what advertising caused”

A click is an event, not a counterfactual.

Click data is useful. It reveals which links people used, which queries generated traffic, which creative executions attracted response and where a customer journey appeared to pass. What it cannot establish by itself is whether the journey required the advertising.

Branded search demonstrates the problem. A person searching a company's exact name is already expressing strong intent. If the paid result receives the click and the platform records the sale, a last-click model can credit the ad with revenue that might have arrived through the organic result anyway. The eBay field experiments are important because they showed how large that gap could be for a major advertiser in a particular setting.

The correction matters because digital advertising's promise of measurability encouraged a category error. Better observation of behaviour after exposure is valuable. Causal measurement still needs a credible comparison with no exposure.

“Brand advertising and direct response are opposites”

They are different jobs on a continuum, often performed by the same campaign.

Direct-response advertising asks for a measurable action soon: buy, register, enquire, download, call. Brand advertising often aims to change future probability by building recognition, memory and expectations before the purchase occasion. The distinction is useful because the appropriate creative, media and measurement differ.

The categories become misleading when they are treated as rival philosophies. A response ad works better when people already recognise and trust the name. A broad campaign can include a direct route to action. A retailer can run television for fame and search for demand capture. The same person can encounter both in one purchase journey.

The correction matters because organisations sometimes force all advertising into the metric of the easiest-to-measure channel. That can starve long-term memory building. The opposite error is to label unaccountable spending "brand" and exempt it from evidence. Different horizons still require economic discipline.

In practice the two modes can reinforce each other. Broad advertising can make later search and sales conversion easier because the name is already familiar. Response data can reveal which claims or audiences deserve broader investment. Treating the modes as connected stages produces better questions than forcing every pound into one camp.

“People hate advertising, so it cannot work”

People often dislike interruption, repetition, tracking and irrelevant commercial noise. They can also enjoy, use or share particular ads. More important, liking is not the same as influence.

An advertisement can be useful when it supplies information at the right moment. It can be entertaining enough to earn attention. It can irritate someone while making the brand easier to retrieve. Negative reactions can also damage the advertiser, especially when repetition becomes harassment or the message violates expectations about privacy and context.

The industry therefore faces a design constraint rather than a yes-or-no verdict. Advertising competes for attention it has not been invited to occupy. It must earn enough value from relevance, information or entertainment to overcome the cost of interruption.

The correction matters because contempt for the audience is commercially expensive. People have avoidance tools, subscription alternatives and public channels for punishing brands. The power relationship is real but incomplete.

“Digital advertising removed the waste”

Digital advertising moved waste into new places.

The old waste was obvious: a national newspaper reached readers who would never buy, a television spot reached households outside the target, a poster could not report who looked. Digital systems can narrow audiences, vary bids and measure interactions. That is a real improvement.

Yet new waste appears as duplicated reach, accidental clicks, fraud, low-quality inventory, excessive frequency, intermediaries' fees, ads shown to existing customers, optimisation towards easy conversions and attribution of sales that would have occurred anyway. Automation can reduce some inefficiency while scaling other errors at machine speed.

The correction matters because "measurable" and "efficient" are different claims. A bad process instrumented perfectly remains bad. Digital systems also make it easier to optimise locally while missing the system-wide outcome, such as shifting sales between channels rather than creating new demand. The proper comparison is not digital versus analogue purity. It is the incremental value created by the next pound after all media, data and operating costs.

Digital systems do offer one decisive advantage: they can change quickly. Bids, creative variants and budgets can be adjusted in response to evidence. That adaptability is valuable only if the optimisation target is worth pursuing. A machine that learns faster towards the wrong proxy becomes confidently wrong faster. The practical gain from digital advertising is therefore iteration, not omniscience. Faster feedback can improve decisions when the feedback variable is connected to real value.

Use It

Start with the change, not the channel

Before choosing search, television, creators or social media, state what must become more likely. More people must know the name. Existing demand must find the product. A misconception must be corrected. Buyers must remember the brand at renewal. A local launch needs enough awareness within five miles. The wording matters because each job implies a different media and measurement design.

"We need Instagram" is not a strategy. It is a proposed distribution choice before the communication problem has been diagnosed. The advertising discipline is to define the change first, then ask which paid opportunities can produce it.

Write the change as a sentence that can later be falsified. "More adults in Manchester who buy running shoes should recognise our name before the September launch" is stronger than "build awareness" because it identifies population, mental outcome and timing. Precision at the objective stage prevents false precision later in the dashboard.

Separate exposure, memory and behaviour

When a report says a campaign "performed", ask at which layer.

Delivery metrics answer whether the media ran. Attention or view metrics say something about opportunity to process. Brand measures test whether perceptions or memories moved. Behavioural metrics show actions. Financial metrics connect actions to margin and value. Causal tests estimate what portion would not have happened otherwise.

Keeping the layers separate prevents a familiar slide-deck trick in which an impressive early-stage metric silently becomes evidence for a later-stage outcome. An ad can deliver excellent reach and weak sales. That may be a creative failure, a product problem, a time-horizon problem or no failure at all if the job was something else. Diagnose the layer rather than declaring victory or defeat too early.

This also improves agency conversations. A media team can be held accountable for whether contracted delivery occurred. Creative can be evaluated against attention and memory tasks. Commercial leadership owns the economics of product, margin and availability. Causal measurement asks whether the whole system produced incremental change. One metric should not carry every organisational responsibility.

Treat every audience as a hypothesis

Whenever you see a target definition, ask what evidence makes the group commercially useful.

"Women aged 25 to 34" may be a convenient buying category with little connection to the decision. "Visited the pricing page in the last seven days" has stronger behavioural relevance but may select people already near purchase. "Chief financial officers at firms with 200 to 1,000 employees" may be useful in one business market and miss the people who initiate the purchase in another.

The best targeting variable is not necessarily the most personal. Context, geography, timing or category behaviour can outperform elaborate profiles. The discipline is to identify the proxy, its likely error and whether the campaign needs prediction, persuasion or both.

Then ask what the targeting costs. Extra data, narrower delivery and more complex optimisation are not free. A broad contextual buy may be operationally simpler and produce more new customers even if its conversion rate is lower. Efficiency should be judged against incremental value per pound, not the cleanliness of the audience definition.

Make the source easy to retrieve

When evaluating creative, remove the industry's decorative language and ask one hard question: if this encounter leaves one trace, what will point back to the advertiser later?

That may be the name, package, sonic cue, colour, character, product demonstration or an unusually clear association with a buying situation. The principle is especially useful when reviewing attractive work. Cover the logo or remove the pack. Does the execution still belong uniquely to this advertiser, or could a competitor inherit it unchanged?

This is not an argument for larger logos and duller work. It is an argument for creative ownership. Distinctiveness lets a brand benefit from the attention it paid to acquire.

A second test is portability. If the core idea works only after the brand name is added at the end, the execution may not be doing enough commercial work during the encounter. If the product, name or distinctive cue is integrated into the story or demonstration, attribution in memory becomes easier. Creative integration is a memory problem disguised as an aesthetic one.

Demand a missing world before believing return

Any claim that advertising caused sales should trigger the same question: compared with what?

The best answer may be a randomised holdout. Sometimes geography, staggered rollout or another quasi-experimental design is more practical. Sometimes causal precision is impossible and the organisation must combine weaker evidence. What matters is recognising the gap rather than filling it with attribution.

This lens changes budget conversations. A channel showing a high reported return can deserve less money if experiments reveal that it mostly harvests existing demand. A broad channel with weak direct attribution can deserve more if credible testing shows incremental sales. The objective is not to reward the most measurable channel. It is to buy the most valuable change.

Where experiments are impossible, use a ladder of evidence rather than pretending certainty. Look for timing, geography, dose-response patterns, consistent results across methods and plausible mechanisms. State what remains uncertain. The discipline is not "experiment or know nothing"; it is refusing to let weak evidence masquerade as strong evidence.

Follow the money through the medium

When a media product behaves strangely, ask who funds it and what gets monetised.

A subscription publication optimises around retained paying readers. An ad-funded platform must also create inventory and commercially useful signals. An influencer balances audience trust against sponsorship revenue. A retailer with an advertising business may earn money from the sale and from selling suppliers access to shoppers. These incentives do not explain every editorial or product decision, but they set pressures that are easy to miss if you look only at the content.

The same lens applies inside the campaign supply chain. Ask how the agency, platform, affiliate, creator and measurement vendor are paid. Incentive mismatches are often more important than bad intentions.

This lens is especially useful when a metric suddenly becomes fashionable. Ask who benefits if the organisation adopts it. A platform proposing a new optimisation goal may have good evidence that it predicts advertiser value; it also benefits when more spend flows through the system. An independent test is valuable precisely because sellers and buyers can both be competent and still have different interests.

The limits

Advertising is an amplifier with constraints. It cannot make an unavailable product easy to buy. It cannot permanently hide poor quality from repeat customers. It cannot eliminate price differences that matter deeply. It cannot make every person attentive or persuadable. It cannot guarantee that a culturally famous campaign creates profit.

Nor can measurement remove uncertainty. Experiments have sampling error. Holdouts can be contaminated. Brand effects can unfold beyond the test period. Platforms change algorithms and interfaces. Privacy rules restrict observation. Some effects are social and cumulative rather than neatly individual.

Advertising can also produce harms that are not captured by return on investment. Misleading claims can distort decisions. Repeated stereotypes can shape cultural expectations. Tracking can intrude on privacy. High-frequency targeting can become exploitative when aimed at vulnerable people. An economically efficient campaign is not automatically a defensible one.

The one thing to keep

When you see an advertisement, do not begin by asking whether it persuaded you.

Ask what was bought around you. Someone paid for access to this moment, chose or inferred an audience, compressed a commercial problem into a message and selected a metric that would later stand in for success. The message may leave nothing. It may leave a name, a feeling, a fact or a route to action. The seller may then claim credit for what happened next.

That sequence is the durable mental model: distribution, memory, action, counterfactual. Advertising changes minds inside that chain, and the industry became powerful by industrialising every link. Once media learned that the chain itself could fund them, the advertisement stopped being a thing placed in the environment and became one of the forces designing the environment.

Terms

Ad auction. A mechanism that allocates advertising opportunities among eligible buyers, often using bids plus predicted quality or relevance rather than price alone. Search and social platforms run enormous numbers of these allocations automatically.

Ad exchange. A digital marketplace through which advertising inventory can be offered to buyers, commonly as part of programmatic trading. Exchanges connect systems rather than representing the whole supply chain.

Advertiser. The organisation paying to communicate. The advertiser may be a business, charity, government body, political campaign or other institution, though this book focuses on commercial advertising.

Agency. An intermediary providing services such as strategy, creative development, media planning, buying, production or measurement. Modern agencies range from full-service groups to highly specialised firms.

Attribution. A rule or model assigning credit for an outcome among observed marketing contacts. Attribution describes or allocates credit; it does not by itself prove causal effect.

Brand advertising. Paid communication aimed mainly at future demand by building recognition, memory, associations or expectations rather than requesting an immediate measurable response.

Call to action. The explicit next step an advertisement asks the audience to take, such as buy, register, call, visit or download.

Click-through rate. Clicks divided by recorded ad impressions. Useful for diagnosing response to some digital formats, but not a general measure of advertising effectiveness.

Contextual targeting. Selecting advertising based on the content or setting in which the ad appears rather than primarily on personal behavioural profiles.

Conversion. A defined action counted as valuable, such as a purchase, lead, registration or app installation. What counts as a conversion is chosen by the advertiser or platform.

Cost per acquisition. Advertising cost divided by attributed acquisitions. It becomes economically meaningful only when the acquisition definition, attribution rule and customer value are sensible.

CPM. Cost per thousand impressions, a standard way of pricing or comparing media delivery. The M comes from the Latin mille, meaning thousand.

Creative. The message as executed: copy, images, film, sound, design and format. In industry usage, "creative" can refer to both the work and the people producing it.

Demand-side platform. Software used by advertisers or agencies to buy digital advertising inventory across sources according to campaign rules and bids.

Direct response. Advertising designed to generate a near-term action that can be recorded, such as an order, enquiry, call or click.

Distinctive asset. A recognisable non-name cue, such as a colour combination, shape, character or sound, that helps people identify a brand and retrieve it from memory.

Frequency. The number of advertising exposures received by a person or audience over a period. Average frequency can hide people receiving far more or less than the average.

Holdout. A group deliberately withheld from advertising so its outcomes can be compared with an exposed group. Randomised holdouts are a strong tool for estimating incrementality.

Impression. A recorded delivery of an advertisement according to a medium or platform's counting rules. It does not guarantee human attention.

Incrementality. The amount of outcome caused by advertising above what would have happened without it. Incrementality is the central causal quantity behind return calculations. It can be positive, negligible or, in principle, negative if advertising creates substitution, irritation or other harmful effects.

Media plan. The organised choice of where, when, how often and at what price advertising will run in order to reach the intended audience.

Native advertising. Paid communication designed to resemble or fit the surrounding editorial or platform format. Clear identification matters because visual integration can otherwise obscure commercial intent. Influencer sponsorship and publisher-produced branded content can create related disclosure questions.

Programmatic advertising. Automated buying and selling of digital advertising inventory using software and data. It includes real-time auctions and other automated transaction types. Programmatic describes the transaction machinery, not a single medium or targeting method.

Reach. The number or proportion of distinct people or households given an opportunity to encounter a campaign over a period. Broad reach can be economically valuable even when individual response rates look low because advertising effects accumulate across a large population.

Retargeting. Advertising directed at people who previously interacted with a site, app, product or other defined signal, often to encourage return or completion.

Return on advertising spend. Attributed revenue divided by advertising spend. It is not the same as profit and can be badly misleading when attribution overstates causation.

Supply-side platform. Software used by publishers to manage and sell digital advertising inventory, including access to programmatic demand.

Targeting. Selecting which people, contexts or moments are eligible to receive advertising based on geography, content, data or predictions.

Viewability. A technical measure of whether a digital ad had a defined opportunity to be seen on screen. It is closer to potential exposure than to verified attention. Viewability standards make delivery more comparable, but a viewable impression can still be ignored.

Each of these terms describes one part of the machine. Their value is in keeping categories separate. "Impression" should not silently become "attention"; "conversion" should not silently become "incremental sale"; "target" should not silently become "persuadable person". Much bad advertising analysis begins when a proxy is promoted into the thing it only approximates. Keep the vocabulary strict and the reasoning improves.

Go Deeper

The compact overview

Winston Fletcher, Advertising: A Very Short Introduction (Oxford University Press, 2010). Fletcher spent a career in the industry and gives a concise map of agencies, media, regulation, research and what advertising can plausibly do. It is older than the platform era's current details, but unusually good at separating the institution from both industry mythology and anti-advertising caricature. Read it first if you want the cleanest continuation from this book rather than a manual of campaign tactics. Its strongest chapters also help place advertising inside the wider commercial and regulatory institution, a useful corrective to books that discuss only copywriting or famous campaigns.

The primary voice

Claude C. Hopkins, Scientific Advertising (Lord & Thomas, 1923). Read it as a document from the direct-response tradition rather than as modern science. Coupons, samples, specific claims, tests and traceable orders show how early advertisers tried to make response observable. Many current performance-marketing habits are recognisable here a century before the dashboard. The warning is equally useful: Hopkins's traceable response is still attribution unless the design creates a valid comparison. Read him for disciplined commercial thinking, not for a century-old answer to modern causal inference. The prose is terse, dogmatic and historically revealing. It shows how strongly the desire to make advertising accountable predates digital media.

The argument about how advertising works

Paul Feldwick, The Anatomy of Humbug: How to Think Differently About Advertising (Matador, 2015). Feldwick sets competing theories of advertising beside one another and refuses the comforting idea that the industry has one settled mechanism. It is especially useful after this book because it complicates any neat division between information, persuasion, memory, fame and social meaning. Feldwick is strongest when showing why advertising practitioners have repeatedly mistaken one useful model for the whole mechanism. If you finish this book wanting to challenge the memory-centred account rather than merely extend it, this is the best next step.

The digital critique

Tim Hwang, Subprime Attention Crisis: Advertising and the Time Bomb at the Heart of the Internet (Farrar, Straus and Giroux, 2020). Hwang attacks the confidence placed in digital advertising markets, measurement and attention quality. His argument is deliberately provocative and should be read as a critique rather than consensus, which makes it useful for testing the platform economics described here. Pair it mentally with the experimental literature: the point is not that digital advertising never works, but that a large, automated market can remain difficult to value correctly. It is short enough to read as a deliberately adversarial companion rather than a second textbook.

Notes and Sources

The Whole Thing in One Page and Why You Should Care

The book's definition of advertising as paid persuasive communication with purchased distribution follows the institutional distinction between advertising, broader marketing, branding and direct selling. The queue boundary assigns this title paid communication, attention, targeting, creative strategy, media buying, measurement, platform economics, regulation and industry history.

Alphabet Inc.'s 2025 Form 10-K reports Google advertising revenues across Search and other, YouTube ads and Google Network. Advertising remained well over two thirds of Alphabet's consolidated revenue. Meta Platforms Inc.'s 2025 Form 10-K reports advertising revenue of roughly $196 billion against total revenue of roughly $201 billion, establishing the company's much tighter revenue dependence on advertising. These filings support the funding-model argument, not a claim that every product choice is directly optimised for advertising.

Notes on the seven ideas

Delivery, attention and media planning. The distinction among served impressions, technical viewability, human attention and causal effect reflects standard measurement practice and the fact that each is a different event. Reach and frequency are classical media-planning concepts. No universal optimal frequency is asserted because response depends on category, objective, creative, context and prior exposure.

Memory and familiarity. Robert B. Zajonc's 1968 mere-exposure work established that repeated exposure can alter affect under experimental conditions. The book uses the finding narrowly. It does not claim repetition inevitably increases preference. The broader emphasis on mental availability, memory cues and distinctive assets is consistent with advertising and brand-effects research, including Byron Sharp, while the title boundary leaves full brand equity and identity systems to Branding in a Hurry.

Creative effectiveness. Sara Rosengren, Martin Eisend, Scott Koslow and Micael Dahlén's 2020 meta-analysis found positive relationships between advertising creativity and several outcomes while showing that effectiveness depends on how creativity is defined and connected to the brand and message. The historical contrast among Hopkins, Reeves, Bernbach and Ogilvy is used to show competing creative philosophies, not a sequence of scientific supersession.

Targeting. Avi Goldfarb and Catherine Tucker's research demonstrates that targeting and obtrusiveness interact in online display advertising. Anja Lambrecht and Catherine Tucker show that retargeting effectiveness depends on information specificity and where consumers are in the decision process. These studies support the book's treatment of targeting as conditional rather than magical.

Media markets and intermediaries. The UK Competition and Markets Authority's 2020 Online Platforms and Digital Advertising market study documents the structure, concentration and opacity of major parts of digital advertising. IAB Tech Lab's OpenRTB specification documents one important technical standard used in programmatic markets. Programmatic trading is simplified in the body because implementations and transaction types vary.

Incrementality and measurement. Randall Lewis and Justin Rao's 2015 Quarterly Journal of Economics paper explains why advertising return can be statistically difficult to estimate even in large experiments. Brett Gordon, Florian Zettelmeyer, Neha Bhargava and Dan Chapsky compared observational advertising-measurement methods with Facebook field experiments and found large discrepancies in many cases. Garrett Johnson, Randall Lewis and David Reiley examine the relationship between experimental power and advertising measurement. These works support the claim that causal inference, not event logging, is the hard measurement problem.

Thomas Blake, Chris Nosko and Steven Tadelis's 2015 Econometrica paper reports large-scale eBay paid-search experiments. The book uses the branded-search result to illustrate selection and demand harvesting. It does not generalise eBay's result to all paid search.

Advertising effectiveness. Raj Sethuraman, Gerard Tellis and Richard Briesch's meta-analysis of brand-advertising elasticities supports the treatment of average advertising effects as positive but heterogeneous and often small relative to ordinary sales variation. The manuscript avoids quoting a universal elasticity because category, time horizon, media and model specification matter.

Notes on the operating history

Agencies and print. Duke University's Hartman Center for Sales, Advertising and Marketing History preserves the J. Walter Thompson and N. W. Ayer records and its Emergence of Advertising in America materials. They support the nineteenth-century transition from space brokerage towards broader agency services. Claims of "first agency" vary with definitions, so the body avoids a single absolute first.

Hopkins and response. Claude Hopkins's Scientific Advertising was published by Lord & Thomas in Chicago in 1923. Its emphasis on coupons, tests, samples and traceable response is treated as an early performance-marketing philosophy, not evidence that Hopkins had solved causal inference.

Broadcasting. Library of Congress research guides and collections document radio advertising in the early broadcast period and sponsor involvement in television. WEAF's 1922 sale of airtime is treated as a conventional milestone rather than the uncontested first paid broadcast advertisement under every definition. Library of Congress material on early television notes that programmes were often produced by advertising agencies on behalf of sponsor clients before networks gained greater control.

The creative revolution. The Volkswagen "Think Small" campaign began in 1959 at Doyle Dane Bernbach. The campaign is used as a compact example of the shift towards integrated art direction, wit and a less hectoring relationship with the reader. No claim is made that DDB invented creativity.

The commercial web. The AT&T banner on HotWired in October 1994 is conventionally described by industry histories as the first web banner advertisement. Earlier networked commercial promotions existed, so the body uses "conventionally treated" rather than an absolute first.

Search, social and programmatic systems. Current platform documentation and IAB technical standards support the general description of auctions, automated bidding and programmatic allocation. Exact ranking and optimisation systems are proprietary and change frequently, so the manuscript does not freeze current product details into the model.

Regulation and current claims

United Kingdom advertising rules. The Advertising Standards Authority and Committee of Advertising Practice maintain the CAP and BCAP Codes. Current CAP guidance states that marketing communications must be obviously identifiable as such where required by the Code and that misleading advertising is prohibited under detailed rules, including substantiation requirements for objective claims.

Consumer law. The unfair commercial practices provisions of Part 4 of the Digital Markets, Competition and Consumers Act 2024 apply to commercial practices from 6 April 2025. Competition and Markets Authority guidance updated in November 2025 explains the replacement and updating of the previous Consumer Protection from Unfair Trading Regulations framework.

UK online-advertising privacy. The Information Commissioner's Office's guidance on storage and access technologies, published in final form on 29 April 2026, states that use of storage and access technologies for online advertising requires consent and that advertising uses do not fall within the listed exceptions. The Data (Use and Access) Act 2025 altered parts of the wider privacy and electronic-communications regime, but the current ICO position above governs the statement made in this book.

European Union. European Commission Digital Services Act material states that platforms must provide advertising transparency and that targeted advertising using sensitive personal data is prohibited, with protection against targeted advertising of minors based on personal data. The body states these restrictions at a high level because implementation details differ by service and obligation category.

What People Get Wrong and Use It

The subliminal-advertising correction follows Sheri Broyles's historical review and Johan Karremans, Wolfgang Stroebe and Jasper Claus's 2006 experiment. Vicary's cinema-sales story is treated as discredited. The narrower priming result is described with its pre-existing thirst condition to avoid converting a constrained laboratory effect into a claim of mind control.

The practical lenses are editorial syntheses from the evidence above. "Demand a missing world" follows the experimental literature. "Treat every audience as a hypothesis" follows the separation of targeting from causation. "Follow the money through the medium" is an institutional inference from advertising-funded media and intermediary incentives.

Bibliography

Primary sources, archives and institutional material

Advertising Standards Authority and Committee of Advertising Practice. The UK Advertising Codes. Current online editions, accessed 9 August 2026.

Alphabet Inc. Annual Report for the Fiscal Year Ended 31 December 2025. Form 10-K. United States Securities and Exchange Commission, 2026.

Competition and Markets Authority. Online Platforms and Digital Advertising: Market Study Final Report. London: CMA, 2020.

Competition and Markets Authority. Unfair Commercial Practices. Guidance under the Digital Markets, Competition and Consumers Act 2024. Updated 18 November 2025.

Duke University, David M. Rubenstein Rare Book and Manuscript Library, Hartman Center for Sales, Advertising and Marketing History. Emergence of Advertising in America and J. Walter Thompson and N. W. Ayer collections.

European Commission. Digital Services Act policy and advertising-transparency materials. Current online material, accessed 9 August 2026.

Hopkins, Claude C. Scientific Advertising. Chicago: Lord & Thomas, 1923.

IAB Tech Lab. OpenRTB 2.6 Specification. Current maintained edition, accessed 9 August 2026.

Information Commissioner's Office. Guidance on the Use of Storage and Access Technologies. Final guidance published 29 April 2026.

Library of Congress. Consumer Advertising During the Great Depression: Radio and Broadcast Advertising and television-commercial collections.

Meta Platforms, Inc. Annual Report for the Fiscal Year Ended 31 December 2025. Form 10-K. United States Securities and Exchange Commission, 2026.

Research and modern works

Blake, Thomas, Chris Nosko and Steven Tadelis. “Consumer Heterogeneity and Paid Search Effectiveness: A Large-Scale Field Experiment.” Econometrica 83, no. 1 (2015): 155-174.

Broyles, Sheri J. “Subliminal Advertising and the Perpetual Popularity of Playing to People's Paranoia.” Journal of Consumer Affairs 40, no. 2 (2006): 392-406.

Feldwick, Paul. The Anatomy of Humbug: How to Think Differently About Advertising. Kibworth Beauchamp: Matador, 2015.

Fletcher, Winston. Advertising: A Very Short Introduction. Oxford: Oxford University Press, 2010.

Goldfarb, Avi, and Catherine Tucker. “Online Display Advertising: Targeting and Obtrusiveness.” Marketing Science 30, no. 3 (2011): 389-404.

Gordon, Brett R., Florian Zettelmeyer, Neha Bhargava and Dan Chapsky. “A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook.” Marketing Science 38, no. 2 (2019): 193-225.

Hwang, Tim. Subprime Attention Crisis: Advertising and the Time Bomb at the Heart of the Internet. New York: Farrar, Straus and Giroux, 2020.

Johnson, Garrett A., Randall A. Lewis and David H. Reiley. “When Less Is More: Data and Power in Advertising Experiments.” Marketing Science 36, no. 1 (2017): 43-53.

Karremans, Johan C., Wolfgang Stroebe and Jasper Claus. “Beyond Vicary's Fantasies: The Impact of Subliminal Priming and Brand Choice.” Journal of Experimental Social Psychology 42, no. 6 (2006): 792-798.

Lambrecht, Anja, and Catherine Tucker. “When Does Retargeting Work? Information Specificity in Online Advertising.” Journal of Marketing Research 50, no. 5 (2013): 561-576.

Lewis, Randall A., and Justin M. Rao. “The Unfavorable Economics of Measuring the Returns to Advertising.” Quarterly Journal of Economics 130, no. 4 (2015): 1941-1973.

Ogilvy, David. Ogilvy on Advertising. New York: Crown, 1983.

Rosengren, Sara, Martin Eisend, Scott Koslow and Micael Dahlén. “A Meta-Analysis of When and How Advertising Creativity Works.” Journal of Marketing 84, no. 6 (2020): 39-56.

Sethuraman, Raj, Gerard J. Tellis and Richard A. Briesch. “How Well Does Advertising Work? Generalizations from Meta-Analysis of Brand Advertising Elasticities.” Journal of Marketing Research 48, no. 3 (2011): 457-471.

Sharp, Byron. How Brands Grow: What Marketers Don't Know. Melbourne: Oxford University Press, 2010.

Zajonc, Robert B. “Attitudinal Effects of Mere Exposure.” Journal of Personality and Social Psychology 9, no. 2, part 2 (1968): 1-27.