THE GREAT ATTENTION EXPERIMENT: 4 - THE BUSINESS MODEL CHANGED EVERYTHING
Imagine two companies offering exactly the same digital service.
The first asks you to pay £10 each month.
The second allows you to use it for free.
At first glance, the second company appears more generous.
It has removed the cost.
Anyone can join.
Growth becomes easier.
People who would never pay for the service can still benefit from it.
But the company still requires revenue.
Engineers must be paid.
Servers must operate.
Security must be maintained.
Investors expect a return.
The money therefore has to come from somewhere else.
Advertisers agree to provide it.
In exchange, they receive access to the people using the service.
That decision changes far more than the price.
It changes what the company needs from the user.
The subscription company needs people to believe the service remains worth paying for.
The advertising company needs people to remain available for advertising.
Both need satisfied users.
Both want people to return.
Both may use personalisation, recommendations and experimentation.
Neither model is automatically ethical or unethical.
But they do not create identical incentives.
Under one model, the user pays with money.
Under the other, the user's presence helps generate the money.
That distinction changed the internet.
And once some of the most important technologies in human life became dependent upon advertising, the commercial value of human attention began influencing what those technologies were designed to do.
The Person Using the Product Is Not Always the Customer
In an ordinary transaction, the relationship is relatively easy to see.
You enter a shop.
You choose something.
You pay.
The business earns revenue by giving you something you value enough to purchase.
The person using the product and the person funding the business are usually the same person.
Advertising-funded digital platforms create a more complicated relationship.
The user receives the service.
The advertiser supplies most of the money.
The platform connects them.
That means the company must satisfy two groups simultaneously.
It must give users enough value to attract and retain them.
It must give advertisers enough access, targeting and measurable results to keep them spending.
This does not mean the user stops mattering.
Without users, there is no audience.
Without a useful or entertaining product, people leave.
Without trust, a platform can decline.
The user remains essential.
But the revenue-producing customer is often the advertiser.
That creates an important difference between what the user wants and what the business needs.
A user may want to enter a platform, accomplish one thing and leave.
The business may benefit if they notice five additional things before doing so.
A user may want a quick answer.
The business may benefit from another page view.
A user may want to see photographs from family members.
The business may benefit from placing recommended content and advertising between them.
A user may want to watch one video.
The business may benefit if another begins automatically.
None of this requires the company to dislike its users.
It requires only a commercial system in which additional attention creates additional opportunities to earn revenue.
Different Models Ask Different Questions
Every business model creates a question.
A retailer asks:
How do we persuade someone to buy?
A subscription service asks:
How do we remain valuable enough that someone continues paying?
A transaction platform asks:
How do we make it easy for people to complete more transactions?
An advertising-funded platform asks:
How do we attract, retain and monetise an audience while continuing to deliver value to advertisers?
These questions overlap.
A subscription company still cares about engagement because people who never use a service may cancel it.
A retailer still advertises.
A transaction platform may collect data.
An advertising platform may sell subscriptions.
Modern technology companies increasingly combine several models.
But the dominant source of revenue still matters because it determines which outcomes possess the greatest commercial importance.
Netflix offers a useful comparison.
Its 2025 annual report stated that revenue was primarily derived from monthly membership fees. Advertising and other sources were growing, but remained an immaterial proportion of total revenue during that year. Its core strategy was to improve the member experience, offer compelling content and attract and retain paying members.
Netflix still wants people to watch.
A member who never finds anything worth viewing may cancel.
The company tests prices, plans, recommendations and product features.
Its ad-supported subscription adds another commercial incentive.
Yet when revenue primarily comes from membership fees, another minute watched does not automatically create another advertising impression.
The central economic question remains whether the member believes the overall service justifies the monthly payment.
Compare that with Meta.
Meta's 2025 annual report stated that substantially all of its revenue came from advertising across its family of applications. It explicitly linked user retention, growth, engagement and time spent to the number of advertising impressions it could deliver and therefore to its financial performance.
The same broad activity, using a digital platform, exists within two different commercial structures.
One primarily monetises continued membership.
The other primarily monetises advertising opportunities created through continued use.
That difference does not determine every design decision.
It changes the direction of the pressure.
The Numbers Reveal the System
Corporate annual reports are not moral confessions.
They are financial documents.
They explain how businesses make money, which measures matter and what risks could threaten future performance.
Read carefully, they reveal the mechanics of the attention economy without requiring anyone to make accusations about private motives.
Meta reported $196.2 billion in advertising revenue during 2025.
Its family of applications averaged 3.58 billion daily active people in December of that year.
The number of advertising impressions delivered increased by 12 per cent during 2025, while the average price per advertisement increased by 9 per cent. Meta said the growth in impressions was driven partly by increases in users and their engagement with its products.
Alphabet reported $294.7 billion in Google advertising revenue during 2025. More than 70 per cent of Alphabet's total revenue came from online advertising.
Snap reported 474 million daily active users during the final quarter of 2025 and an average quarterly revenue per user of $3.62. Advertising still represented approximately 87 per cent of its annual revenue, despite the company developing subscription and partnership income.
These figures do not prove that the companies deliberately harm attention.
They reveal which relationships are economically central.
Users create activity.
Activity creates advertising opportunities.
Advertising opportunities generate revenue.
Advertisers continue spending when the platform can demonstrate results.
The system therefore needs several things to happen at once.
More people must join.
Existing users must return.
They must remain engaged.
Enough advertising must be displayed.
The advertisements must be sufficiently relevant and effective.
Advertisers must believe the return is competitive with other places they could spend their money.
The entire machine depends upon the continued presence and responsiveness of human beings.
Engagement Began as a Sensible Measure
There is nothing inherently sinister about measuring engagement.
A company needs to know whether people find its product useful.
If users join and never return, something is wrong.
If a new feature is ignored, resources may be better invested elsewhere.
If people abandon a page because it loads slowly, engineers should fix it.
If a safety feature reduces abuse, the company should know.
Engagement can indicate value.
A meaningful conversation is engagement.
Learning something is engagement.
Watching a video from a distant family member is engagement.
Finding a community is engagement.
Creating art, sharing knowledge and organising support can all be measured through activity.
The problem is not the existence of an engagement metric.
The problem is the assumption that more engagement always represents more value.
A person may remain because the experience is useful.
They may also remain because it is difficult to stop.
They may return because a relationship matters.
They may return because a notification created anxiety.
They may watch another video because it is fascinating.
They may watch because the next one started before they consciously decided whether they wanted it.
A metric records the behaviour.
It does not automatically explain the human meaning behind it.
Ten minutes of conversation with a close friend and ten minutes of compulsive scrolling can both appear as time spent.
A return visit prompted by genuine interest and one prompted by fear of missing out can both appear as retention.
A share intended to help someone and a share driven by outrage can both appear as engagement.
The number rises.
The quality of the experience remains uncertain.
What Gets Measured Begins to Shape What Gets Built
Metrics do not simply report what an organisation has done.
They influence what it does next.
Imagine a product team testing two versions of a feature.
Version A helps users complete their intended task quickly and leave.
Version B encourages them to remain for an additional four minutes.
If the team is measured primarily through time spent, Version B appears to win.
Now imagine another test.
Version A produces thoughtful interactions among a smaller number of people.
Version B generates twice as many reactions through emotionally provocative content.
If total interactions are the central measure, Version B may appear more successful.
This does not mean product teams are incapable of understanding nuance.
Large companies measure safety, satisfaction, relevance, quality and many other outcomes.
Different metrics can be balanced.
Human judgement remains part of the process.
But the metric most closely connected to revenue carries particular weight.
A company can sincerely value wellbeing while operating a model in which reduced use may reduce income.
That tension exists before anyone makes a decision.
The business model places it there.
The Product Became a Continuous Experiment
Traditional products changed slowly.
A manufacturer designed something.
Produced it.
Distributed it.
Waited for sales figures, customer complaints and market research.
A new version might appear months or years later.
Digital products can change continuously.
Two people may use what appears to be the same application while encountering slightly different versions.
One button is larger.
One notification uses different wording.
One feed contains more recommended material.
One group sees a feature another group does not.
The company compares the outcomes.
Which group clicked more?
Which remained longer?
Which returned sooner?
Which completed the desired action?
The more successful version can then be distributed more widely.
Meta described this process publicly as early as 2012, explaining that product development involved continuous iteration, testing and measuring how people responded.
In 2014, Facebook described Airlock, its mobile A/B-testing framework. The system allowed different users to receive different versions of an application so that the company could compare metric data and decide which version to release or how to continue developing it.
Experimentation itself is not wrong.
It can improve accessibility.
Reduce loading times.
Identify bugs.
Make privacy settings clearer.
Decrease harmful content.
Improve safety.
Find features people genuinely value.
Any responsible digital company should test whether its products work.
But the outcome chosen to define success matters enormously.
If the test asks which version helps users accomplish their goal, the product moves in one direction.
If it asks which version creates more advertising opportunities, it may move in another.
If it asks both questions, the company must decide what happens when the answers conflict.
Tiny Changes Become Enormous at Scale
A two per cent increase sounds small.
Inside a product used by billions of people, it can be extraordinary.
Suppose a design change encourages the average person to remain for thirty additional seconds.
For one user, the difference is barely noticeable.
Across one billion users, it represents more than fifteen million additional hours of collective attention.
If some proportion of that time creates advertising impressions, the commercial value can be substantial.
This is one reason digital products become highly optimised.
A minor improvement in retention can affect millions of people.
A slight increase in click-through rate can generate enormous revenue.
A tiny reduction in the number of people who close an application can influence quarterly results.
At scale, fractions become fortunes.
The user experiences one changed button.
The company experiences a measurable movement across an entire population.
The asymmetry is extraordinary.
An individual cannot easily detect the effect of every design choice upon their own behaviour.
The company can compare millions of interactions and identify patterns invisible to any single person.
This does not make the company omniscient.
Experiments can produce misleading results.
Metrics can be noisy.
Users behave differently across cultures and contexts.
Short-term gains may create long-term losses.
But the company still possesses a level of behavioural visibility no ordinary user can match.
The product learns from the population.
The individual sees only the version that reaches them.
The Flywheel
Advertising-funded platforms can develop a self-reinforcing cycle.
More users make the service more useful.
A social network becomes more valuable when friends, family members, creators and organisations are already there.
A video platform becomes more useful as more people upload content.
A search engine improves as it processes more queries and observes which results satisfy them.
More users create more activity.
More activity creates more information.
More information can improve recommendations and advertising.
Better targeting can make the platform more valuable to advertisers.
More advertiser spending provides more money for infrastructure, acquisitions, research and product development.
Those investments can attract more users.
The cycle begins again.
Economists describe parts of this process as network effects.
The UK's Competition and Markets Authority found that social-media platforms can benefit from cross-side network effects between users, content providers and advertisers. More users can attract more content and advertisers, while more content can attract more users.
The cycle creates enormous value.
It helps explain why platforms can offer sophisticated services without direct payment from most users.
It also creates barriers to meaningful competition.
A new social network does not merely need a better application.
It needs the people someone already knows.
A new video service needs content.
Creators go where the audience exists.
Advertisers go where they can reach the audience.
Data and revenue concentrate where previous success has already produced scale.
The largest systems therefore gain more than market share.
They gain a growing ability to observe what attracts attention and to invest in becoming better at retaining it.
Growth Changes the Meaning of Success
A small social network can celebrate reaching its first thousand users.
A global platform cannot continue satisfying investors merely by remaining the same size.
Public companies are expected to report financial performance.
Analysts produce forecasts.
Share prices respond to growth, revenue, profit, user numbers and expectations about the future.
This does not mean every publicly listed company must maximise short-term profit at any cost.
It does not mean executives possess no freedom to invest for the long term.
Many technology companies have dual-class share structures that give founders significant control over strategic decisions.
Meta's own filings state that Mark Zuckerberg controls a majority of the company's voting power. Snap's filings say its co-founders control almost all stockholder voting power.
The simplistic claim that anonymous shareholders directly dictate every product decision would therefore be inaccurate.
But public-market expectations still matter.
Companies report active users.
Engagement.
Average revenue per user.
Advertising impressions.
Revenue growth.
Operating margins.
They identify declines in those measures as risks.
Snap states that fluctuations in user growth, retention and engagement, along with failure to meet investor expectations, can affect its market value.
Meta warns investors that lower user retention, growth or engagement could reduce advertising impressions and materially damage its revenue and financial performance.
Once these measures become part of how a company explains itself to the financial world, they are not merely internal statistics.
They become promises.
A quarter of slower growth requires explanation.
A decline in use can affect confidence.
A competitor capturing more time from younger users becomes a commercial threat.
A product decision that reduces engagement may improve lives while creating a financial cost the company must absorb and justify.
The pressure is not a conspiracy.
It is visible in the reporting structure.
The User Becomes a Unit of Economic Performance
Snap reports average revenue per user.
Meta has historically reported average revenue per person.
These are perfectly rational business measures.
Companies need to understand how effectively they monetise their audiences.
But the language reveals what the business model does.
A human being becomes both a person using the service and a unit through which revenue is calculated.
Their activity contributes to the denominator.
Advertising income contributes to the numerator.
The company can then compare regions.
One user population may generate more revenue than another.
One format may monetise more effectively.
One product may produce more advertising impressions.
One behaviour may be more commercially useful.
This does not mean executives cease seeing users as human beings.
It means the financial system must abstract them.
Businesses cannot report the individual meaning created for billions of people.
They report numbers.
Daily active people.
Monthly active people.
Time spent.
Impressions.
Clicks.
Average revenue.
Retention.
The abstraction is necessary for management.
It also risks turning the measurable part of the relationship into the only part that appears real.
The Business Model Does Not Need to Demand Addiction
A company does not need to write make this addictive into a product brief.
That language would attract scrutiny.
It is also unnecessary.
The company can ask teams to increase retention.
Improve daily use.
Reduce abandonment.
Increase video completion.
Grow advertising inventory.
Raise average revenue per user.
Encourage sharing.
Strengthen notifications.
Improve recommendations.
Each instruction sounds ordinary.
Each can be justified through business logic.
Together, they may create a product that becomes increasingly difficult to leave.
This is why focusing only on malicious intent misses the larger problem.
Systems do not need immoral instructions to produce harmful outcomes.
They need incentives that reward one measurable behaviour while ignoring costs appearing elsewhere.
A product team may improve its target.
The company may increase revenue.
The user may lose another twenty minutes each day.
The child's bedtime may move later.
The conversation may be interrupted.
The work may become fragmented.
None of those consequences necessarily appear inside the experiment dashboard.
The metric can improve while the life surrounding it worsens.
The Difference Between Satisfaction and Consumption
A good business wants satisfied customers.
An attention business also benefits from consumption.
These are not identical.
Imagine eating in a restaurant.
A satisfying meal eventually creates fullness.
The customer stops.
The stopping point is part of the experience.
Now imagine a restaurant whose revenue increases with every additional bite but where the customer never feels full.
The commercial incentive would change.
Portions would not need natural endings.
Flavours might become more intense.
The next bite could arrive automatically.
The restaurant could still serve enjoyable food.
Its interest in satisfaction would coexist with an interest in continued consumption.
Digital attention works in a similar way.
A platform may genuinely want users to enjoy themselves.
Enjoyment supports retention.
But satisfaction that causes someone to close the application creates fewer immediate opportunities than satisfaction that encourages them to continue.
The ideal commercial experience becomes one that delivers enough value to prevent abandonment without producing a clear feeling of completion.
This is why natural endings matter.
A newspaper has a final page.
A television programme reaches credits.
A book closes.
An endless feed has no equivalent.
The absence of an ending is not only a design choice.
Inside an advertising model, it has economic value.
The specific psychological techniques through which products encourage continued use will be examined in the next essay.
Here, the important point is simpler.
The business model creates a reason to remove stopping points.
Not Every Advertising Business Needs Endless Attention
This argument requires an important qualification.
Advertising-funded businesses do not all benefit from maximising time spent in exactly the same way.
Search advertising can succeed when someone arrives, expresses an intention, clicks a useful result and leaves.
A search engine may create enormous commercial value in a brief interaction.
Classified advertising can connect a buyer and seller quickly.
A map can display a sponsored result while helping someone reach a destination.
An advertising-funded email service does not necessarily need someone to stare at their inbox all day.
The relationship between attention and revenue differs across products.
It would therefore be inaccurate to claim that every advertising company wants every user to remain indefinitely.
What is true is that advertising businesses need opportunities to present and measure advertisements.
Some create those opportunities through frequent short visits.
Others through long sessions.
Others through high-intent searches.
Others through repeated exposure across an ecosystem.
The form changes.
The dependency remains.
More commercially useful activity generally creates more opportunity than no activity at all.
When Safety and Revenue Pull in Different Directions
The most revealing evidence of the business model is not that companies value engagement.
That is obvious.
It is what happens when another objective may reduce it.
Meta's 2025 annual report acknowledges that changes intended to improve privacy, safety, security or age-appropriate experiences can reduce time spent, engagement and monetisation opportunities.
The company states that updating feed-ranking systems or product features to improve user experience can reduce some measures of engagement and adversely affect financial results. It also notes that enforcing policies designed to protect security and platform integrity can reduce advertising revenue.
This is an extraordinarily important admission.
Not because it proves that Meta refuses to make protective changes.
The same filing confirms that the company does make them.
Not because every safety intervention reduces revenue.
Many may strengthen trust and improve long-term performance.
The significance is that the conflict exists.
A decision can be better for privacy, safety or age-appropriate design while creating a commercial disadvantage.
The company must then decide how much disadvantage it is prepared to accept.
That is where moral obligation enters the business model.
An organisation funded through engagement may sometimes have to choose actions that reduce the very metric upon which its revenue depends.
The existence of that choice does not prove how it will be resolved.
It proves that goodwill alone cannot eliminate the tension.
Privacy Can Carry a Commercial Cost
The same tension appears around data.
Meta states that restrictions introduced by regulators, browsers, mobile operating systems and its own product changes have reduced its ability to target and measure advertising effectively.
It describes these limitations as harmful to advertising revenue and potentially to engagement.
Again, this does not prove that privacy should always override every other consideration.
Personalisation can improve relevance.
Measurement can help businesses avoid wasting money.
Smaller advertisers may depend upon effective targeting.
But the financial consequence matters.
A company whose revenue depends upon data-driven advertising has an economic reason to resist changes that make targeting less precise.
That does not mean every objection is dishonest.
It means the company cannot approach the question as a neutral observer.
Its interpretation of privacy, measurement and targeting takes place inside a model that profits from access to information.
The same applies to safety.
And time spent.
And age verification.
And limits upon personalisation.
Every protective intervention enters a commercial system with existing dependencies.
Good People Still Operate Inside Incentives
There are tens of thousands of people working inside large technology companies.
Engineers.
Researchers.
Designers.
Safety specialists.
Policy teams.
Moderators.
Parents.
People who genuinely care about the effects of what they build.
Treating them as a single malicious organism would be ridiculous.
Many have spent years improving privacy, accessibility, safety and user control.
Some have challenged their own employers.
Some have left.
Some believe the benefits still outweigh the harms.
Some are attempting to change products from within.
The problem is not that every person inside the system lacks integrity.
The problem is that integrity must repeatedly negotiate with incentives.
An employee may want to introduce more friction before a potentially harmful action.
The business may worry that friction reduces use.
A safety team may want stricter enforcement.
Revenue teams may see a loss of advertising inventory.
A researcher may identify a long-term risk.
Product leadership may be measured against a short-term target.
An executive may sincerely believe a change will help users while also knowing it improves retention.
Real decisions rarely divide cleanly into good and evil.
That complexity does not remove responsibility.
It explains why the outcome cannot depend solely upon personal virtue.
A system that requires exceptional individuals to resist its normal incentives has been designed badly.
The Mission and the Model
Many technology companies began with missions larger than profit.
Organise the world's information.
Connect people.
Give everyone a voice.
Build community.
Encourage creativity.
Those ambitions were not necessarily fraudulent.
The services delivered real progress towards them.
But a mission describes what an organisation wants to contribute.
A business model determines how it survives.
When the two align, growth can strengthen the mission.
More users create more connection.
More revenue funds better infrastructure.
Better tools expand access.
The relationship feels virtuous.
The difficulty begins when serving the mission requires reducing something the model rewards.
Less engagement.
Fewer advertisements.
Less data.
More friction.
Slower growth.
Stronger limits for children.
A mission statement cannot resolve that conflict.
It can guide the decision.
It cannot remove the cost.
This is why business models matter more than corporate slogans.
A company may genuinely believe in its mission.
Its incentives still shape which ideas receive investment, which experiments count as successful and which compromises become difficult to make.
The Moral Turning Point
Nobody meant to build the attention economy.
The first essay in this series established that.
Useful services needed funding.
Advertising provided it.
Engagement indicated value.
Personalisation made products more relevant.
Experimentation improved them.
Growth expanded access.
Each step made sense.
But accidental beginnings do not guarantee permanent innocence.
Once a company understands that its system may contribute to harm, the question changes.
It is no longer whether the original founders intended the outcome.
It is whether the business is prepared to sacrifice revenue, growth or competitive advantage to reduce it.
That question cannot be answered through a mission statement.
It must be answered through decisions.
Which metrics are changed?
Which features are redesigned?
Which protections are introduced?
Which users are excluded from certain forms of targeting?
Which recommendations are limited?
How much friction is accepted?
How much revenue is surrendered?
How quickly does the company act?
What happens when a competitor refuses to make the same compromise?
The evidence concerning what companies knew, how they responded and whether financial incentives overrode moral obligations will be examined later in this project.
Essay Four establishes why that evidence matters.
The business model ensures that meaningful protection may carry a price.
Regulation Is an Attempt to Change the Calculation
Calls for regulation are often described as government interference with innovation.
Sometimes they are.
Poorly written rules can entrench large companies, damage useful services and impose costs smaller competitors cannot absorb.
Regulation can lag behind technology.
It can target yesterday's problem.
It can create unintended consequences of its own.
But the basic purpose of regulation is to alter incentives when the market does not account adequately for wider costs.
Pollution is the obvious example.
A factory may create a useful product and profitable employment while pushing environmental costs onto everyone living nearby.
If the business does not pay those costs, the market price of the product conceals part of the truth.
Attention can create similar external costs.
The platform receives revenue.
The advertiser receives access.
The user receives a service.
The costs of interrupted sleep, reduced concentration, family conflict, educational disruption or deteriorating wellbeing may be experienced elsewhere.
By children.
Parents.
Schools.
Employers.
Healthcare systems.
Communities.
If those costs do not appear on the company's financial statement, the business model has little automatic reason to reduce them.
Regulation, design standards and legal duties attempt to move some of those costs back into the commercial calculation.
Whether current interventions achieve that effectively remains an open question.
Could the Model Be Different?
The obvious alternative is to charge users directly.
Subscriptions reduce dependence upon advertising.
They can also exclude people who cannot afford them.
A world in which education, communication and information sit behind multiple monthly payments would create its own inequality.
Micropayments have been proposed for decades but remain inconvenient and unpopular.
Publicly funded digital infrastructure raises questions about political control.
Non-profit models depend upon donations or grants.
Transactional models work for some services and not others.
Contextual advertising reduces the need for behavioural profiling but may generate less revenue for certain businesses.
Hybrid models combine subscriptions, advertising, commerce and services, but their incentives remain mixed.
There is no perfect model waiting to replace the current system.
Every source of revenue creates pressure.
The task is not to find a business with no incentive to influence behaviour.
It is to align commercial success more closely with genuine human value.
That may mean subscriptions in some contexts.
Contextual rather than behavioural advertising in others.
Stronger default protections for children.
Metrics that include satisfaction and long-term wellbeing.
Restrictions upon particular forms of targeting.
Independent auditing.
Greater transparency.
Products designed around clear stopping points.
Alternative ownership structures.
Regulation that rewards safer design rather than simply punishing scandal after it occurs.
The precise solutions deserve their own examination.
The first step is recognising that the current outcomes are not produced by technology alone.
They are produced by the way the technology is funded.
The Question Behind Every Feature
Whenever a digital platform introduces a feature, we tend to ask whether it is useful.
Perhaps we should also ask:
What does the business model reward this feature for doing?
Does it help the user complete something?
Does it produce another advertising impression?
Does it gather more data?
Does it encourage a return?
Does it make leaving less likely?
Does it increase the number of people who pay?
Does it strengthen a network effect?
Does it move activity into a part of the product that generates more revenue?
The answers may overlap.
A feature can genuinely improve the user experience while also increasing monetisation.
That is often how successful product development works.
The difficult questions arise where the benefits separate.
When the feature generates more use but less value.
More reaction but less understanding.
More data but less privacy.
More commercial opportunity but less agency.
The business model does not dictate the answer.
It determines which answer is easiest to justify.
Final Thought
The internet did not become an attention economy merely because technology improved.
It became an attention economy because human attention became connected to revenue.
Once advertising paid for the services, engagement stopped being only evidence that people valued a product.
It became part of the product's commercial value.
More users created more opportunities.
More time created more inventory.
More behaviour created more data.
More data improved targeting.
Better targeting attracted more advertiser spending.
The money funded better systems for attracting, measuring and retaining attention.
A cycle formed.
Nobody needed to order the industry to maximise distraction.
They needed only to measure engagement, reward growth and connect both to revenue.
That is why the business model changed everything.
It changed what success looked like.
It changed which metrics mattered.
It changed what experimentation was designed to discover.
It changed what companies stood to lose when people spent less time on their products.
It created a world in which protecting human attention could conflict with the financial systems built around capturing it.
The question is no longer whether the companies originally intended that outcome.
Most did not.
The question is what happens when protecting the user requires the business to accept less of the thing that made it successful.
Less attention.
Less data.
Less growth.
Less money.
That is where values stop being words.
And become decisions.
Sources & Research Gaps
Principal Sources
Meta Platforms, 2025 Annual Report
Meta's annual report provides primary evidence concerning its dependence upon advertising, the relationship between engagement and advertising impressions, daily active people, advertising revenue, targeting, measurement and the potential financial effects of privacy, safety and user-experience changes.
The company reported:
- $196.2 billion in advertising revenue during 2025;
- 3.58 billion average daily active people in December 2025;
- a 12 per cent increase in advertising impressions during 2025;
- a 9 per cent increase in average price per advertisement;
- that reduced retention, growth, time spent or engagement could harm advertising revenue;
- that some changes intended to improve privacy, safety, integrity and age-appropriate experiences may reduce engagement or monetisation.
Alphabet, 2025 Annual Report
Alphabet's filing states that more than 70 per cent of total revenue came from online advertising in 2025.
It reported $294.7 billion in Google advertising revenue across Search, YouTube and the Google Network.
Snap, 2025 Annual Report
Snap's filing identifies daily active users as a critical engagement measure and average revenue per user as a key commercial metric.
The company reported:
- 474 million daily active users during the final quarter of 2025;
- quarterly average revenue per user of $3.62;
- $5.9 billion in annual revenue;
- approximately 87 per cent of annual revenue from advertising;
- that changes in user growth, retention, engagement and investor expectations can affect financial performance and market value.
Netflix, 2025 Annual Report
Netflix provides a useful comparison with a predominantly subscription-funded business.
Its filing states that revenue was primarily derived from monthly membership fees. Although advertising revenue was growing, non-membership revenue was not yet a material component of total revenue during 2025.
The filing also shows that subscription companies still care about engagement, satisfaction, retention, pricing and experimentation.
Meta Engineering, Building and Testing at Facebook
Meta's engineering publications describe the company's approach to continuous iteration and product testing.
The Airlock mobile A/B-testing framework allowed different groups of users to receive different versions of an application so that product teams could compare performance metrics and determine which version to release.
Facebook 2012 Registration Statement and Annual Reports
Facebook's earlier filings demonstrate that the relationship between activity, advertising opportunities and revenue was visible from the company's early years as a public business.
Its registration statement warned that user activity taking place outside Facebook could reduce advertising opportunities and revenue.
Later filings connected user growth, engagement, advertisements delivered and advertising prices directly to commercial performance.
UK Competition and Markets Authority
The CMA's work on online platforms and digital advertising describes cross-side network effects between platform users, content providers and advertisers.
These effects can create self-reinforcing advantages for established platforms and make competition more difficult.
Research Gaps and Limitations
This essay compares broad business-model incentives. Individual companies operate differently, even where they share similar revenue sources.
Advertising-funded products do not all benefit equally from increased time spent.
Search engines can create commercial value through short, high-intent interactions. Social, entertainment and video platforms may depend more heavily upon repeated or prolonged use.
Subscription models are not inherently protective of attention.
They still benefit from retention, habit formation and strong engagement. Some also include advertising, in-app purchases, data collection or other monetisation systems.
Corporate annual reports identify commercial risks and dependencies. They do not reveal every internal product decision, debate or ethical consideration.
The inclusion of engagement as a business metric does not prove that a company seeks engagement at any cost.
A/B testing is a neutral method.
It can be used to increase revenue, improve accessibility, strengthen safety, reduce harmful content or increase user satisfaction.
Evidence that privacy or safety changes may reduce revenue does not prove that a company refused to make those changes.
It establishes that the financial conflict exists.
Public companies face expectations from investors and analysts, but executives retain strategic discretion. Founder-controlled companies may possess greater freedom than businesses with more dispersed voting power.
The legal duties of directors do not generally amount to a simple requirement to maximise short-term profit. Corporate governance, long-term strategy, stakeholder interests and founder control create a more complicated picture.
Further research is needed into:
- which metrics carry the greatest influence in individual product decisions;
- how companies balance engagement against safety, satisfaction and wellbeing;
- how executive and employee compensation relates to user-growth or revenue targets;
- the effects of hybrid advertising and subscription models;
- whether contextual advertising could support major free services at comparable scale;
- how changes in engagement translate into advertising revenue across different products;
- how often experiments optimise for long-term user outcomes rather than short-term activity;
- whether wellbeing metrics can be measured reliably without creating new forms of surveillance;
- the commercial effects of age-appropriate design and stronger protections for children;
- the internal decisions companies make when safety interventions reduce growth or revenue.
This essay establishes the structural incentive.
Later essays will examine the psychological techniques used to increase engagement and the evidence concerning what companies knew about possible harms.