Note : This article outlines the fundamentals of modern online advertising and algorithm-driven systems. While not comprehensive, it aims to give aspiring online entrepreneurs particularly those with some technical background, a clearer view of how these models operate in practice, from free games to mobile applications and everyday online services.
Advertisers don’t usually get money directly from viewers. Surprised?
They make money indirectly, in several reinforcing ways even if you never buy or click on anything or you skip the ad later.
Here’s how the system really works.
1. THEY ARE PAYING FOR ATTENTION AND NOT PURCHASES
Most ads are bought using models like:
CPM (cost per 1,000 views)
CPC (cost per click)
CPV (cost per video view)
So when you view an ad:
The platform (YouTube, Facebook, Google, TikTok) gets paid
The advertiser is spending, not earning
At this stage, you’re the product, not the customer.
Behind the scenes, ad-targeting algorithms assess your likelihood of engaging. Even if you never click, the system predicts value based on behaviour not intent.
2. BRAND AWARENESS IS THE FUTURE MONEY
Even if you don’t buy now, the "sub-conscious" mind will make :
You recognise the brand later,
You subconsciously trust what feels familiar,
When you do need that product, their name pops up first
Classic example:
You’ve seen a brand 20 times, months later you choose it “for no reason”. That “no reason” is advertising doing its quiet, long-term work. Brand advertisers don’t need clicks, just mere exposure is enough.
3. DATA IS BEING COLLECTED
When you view an ad, platforms learn:
Your age group (demographic), interests, habits,
What you pause on,
What you skip instantly,
What people like you eventually buy
Importantly, this includes non-actions, skipping, muting, refreshing, or avoiding ads altogether.
How do they know this? Well, most apps are signed in with your email. You feel no pressure but you are already giving vital information to them.
That data:
Improves future targeting,
Makes ads cheaper and more effective,
Feeds machine-learning models that predict behaviour
You didn’t buy, but your behaviour trained the algorithm.
4. RETARGETING : TODAY'S VIEW IS TOMMORROW'S SALE
Ever noticed:
You saw an ad once. Then it followed you everywhere?
That’s because:
Your view put you into a “warm audience”,
Advertisers pay more for you later because you’re statistically more likely to convert,
Algorithms then find lookalike users, people whose behaviour resembles yours multiplying the value of a single unpaid view.
One skipped ad can still lead to a higher-value customer elsewhere down the line.
5. NETWORK EFFECTS AND SOCIAL PROOF
High view counts do this:
“Everyone is talking about this”,
“This must be popular”
“I keep seeing it, so it must be legit”
Popularity itself becomes a selling tool even if most viewers never buy. At scale, perception is market power.
6. SOME ADS AREN'T SELLING TO YOU AT ALL
Many ads are designed to:
Impress investors,
Attract distributors,
Recruit talent,
Signal dominance and legitimacy
Your view helps them look bigger, established, and credible even if you never engage.
7. THE HONEST TRUTH
If you:
Watch ads,
Don’t buy,
Skip quickly
You still create value just not cash in your pocket. Advertising today is less about convincing individuals and more about shaping markets, habits, and perception over time.
8. HOW THE ALGORITHMS DECIDE ALL THESE
There isn’t one algorithm but there’s a stack of ML/AI models working together. Here are some examples : (There's more actually, but read on...)
8.1 Ad targeting algorithm (behaviour-based) : This decides who should see an ad.
It tracks:
Clicks, views, likes
Watch time, skips, searches,
Even patterns of avoidance
Users are classified into probabilistic segments (“likely gamer”, “pet owner”, “price-sensitive”, “ad-resistant”).
Machine-learning models such as logistic regression, random forests, XGBoost, and deep neural networks predict engagement, conversion, or retention.
8.2 Bid optimisation algorithm
Every ad slot triggers an auction. Advertisers submit bids (CPM/CPC/CPA).
The system calculates: Expected Value = Bid × Predicted Engagement Probability
The highest expected value wins not necessarily the highest bid. Reinforcement learning continuously adjusts bidding strategies based on outcomes.
8.3 Recommendation and ranking algorithms
Once ads are eligible, platforms decide which ad appears first.
They consider:
Your personal history,
Engagement probability,
Ad relevance
Diversity (to avoid repetition fatigue)
This is why ads increasingly feel “organic” especially on platforms like TikTok.
8.4 Retargeting and lookalike algorithms
If you viewed but didn’t buy - You enter a retargeting pool where similar users are identified using clustering and embedding-based similarity models.
Your single interaction scales across thousands of others.
So, there Is no Single Algorithm, there Is an Algorithmic Stack
Advertising platforms do not rely on one algorithm, but a layered system working simultaneously:
User profiling models predict interests and behaviour patterns,
Auction algorithms decide which advertiser wins an ad slot,
Ranking and recommendation systems choose which ad appears first,
Retargeting and lookalike models find similar users,
Attribution models decide which exposure “deserves credit”
These systems continuously learn not just from clicks, but from avoidance
9. ATTRIBUTION AND CONVERSION PREDICTION
Advertisers still want to know what worked. Models like Markov chains, logistic regression, and survival analysis assign partial credit across multiple views, clicks, and time delays, optimising budgets for ROI, not fairness.
10. TL;DR (Too Long, Didn't Read)
Advertising today runs on prediction, not persuasion.
Targeting models find the user,
Auctions price the attention,
Ranking decides visibility,
Retargeting scales behaviour
Attribution assigns value
Every click, view, skip, or idle hover feeds the system. Even resistance is data.
11. WHO REALLY PAYS FOR "FREE" CONTENT
Rule 101 : There is nothing free on the internet, otherwise companies go busted quickly
In the digital age, we are constantly told that online services are “free.” Free videos, free search, free social media, free content on demand. Yet behind this apparent generosity lies a complex economic system built not on charity, but on attention, data, and prediction.
A common question arises: If users skip ads, never click, and never buy, how do advertisers and platforms still make money? And more importantly,
12. WHAT ABOUT ISPs/TELCOS? Are They Part of the Money Loop?
A persistent rumour suggests that advertisers or platforms also charge Internet Service Providers (ISPs), telcos or vice versa, based on ad consumption. This is largely untrue in direct terms, but partially true in indirect ways.
Platforms do not charge ISPs per ad view,
ISPs do not earn money directly from ads
However:
Platforms pay ISPs for infrastructure, colocation, and traffic delivery,
Ads increase data consumption, driving network expansion,
In some regions, “sponsored data” or zero-rating schemes existed
Telcos increasingly lobby for platforms to pay “network usage fees”
So while ads do not trigger direct ISP billing, traffic economics link them indirectly.
12.1 Why This Feels Like Double Payment
From the user’s perspective:
Users pay the ISP for access,
Users pay platforms with data,
Advertisers pay platforms,
Platforms grow exponentially wealthy
Even without a single illegal transaction, the system feels imbalanced. The real currency is not money alone, it is control over data flows and prediction capability.
13. DOES THIS ADVERTISING SYSTEM INTRUDES ONE'S PRIVACY?
The answers are uncomfortable but worth examining.
The answer is YES (at least not directly) Many people, including regulators and technologists, agree with me. What I am describing does sit in a grey (and sometimes dark) zone between convenience, commerce, and privacy. But it’s a normalized, legalised, and obscured one.
This is where the debate becomes ethical, not just technical but let’s talk about it plainly, not defensively.
Most users:
Don’t fully understand what’s collected
Don’t realise how much can be inferred from “non-actions” (skipping, pausing, hovering)
Don’t meaningfully consent, they accept to proceed,
Behavioural patterns are tracked even without interaction,
Inferences are drawn from what they skip, avoid, or ignore,
Consent is often buried inside unreadable terms and conditions. It's not informed consent in any human sense. It’s procedural consent.
Why platforms say it’s “not a violation”? Their defense usually rests on three pillars:
a. “It’s anonymised” - Technically true but : "Anonymised" doesn't mean "non-identifiable"
Reidentification via pattern matching is well-documented
b. Because “You agreed to the terms” - also technically true but: Terms are unreadable by design
Power imbalance exists (use the service or leave)
c. “It improves user experience” : Sometimes true but:
That benefit primarily serves advertisers
The user benefit is secondary and optional
So "legality" is not the same as "morality"
In short, this model is widely described as surveillance capitalism not surveillance by force, but by convenience.
Why?
You are : Observed, Categorised, Predicted and Monetised. Without a direct transaction or consent.
And crucially: The prediction is more valuable than the person.
So is it llegal?
Is it illegal? Depends where you stand. In places with strong laws:
GDPR (EU) : limits profiling, requires explicit consent,
PDPA (Malaysia) : consent-based, but perhaps weaker on behavioral inference,
CCPA (California) : opt-out model, not opt-in
But enforcement often lags behind technology as the system evolves faster than the law.
Why most people tolerate it?
Because the trade-off feels invisible,
the notion of “Free” services,
Personalised feeds
Surprisingly - convenience with no immediate harm (but death by a thousand frictionless cuts)
Your instinct is actually the critical one. The real issue isn’t ads. It’s this question:
Should behavior itself be harvested as an economic resource without meaningful consent?
That question is now being debated at policy level, ethics boards, AI governance frameworks, even central banks and national security circles
So I am not being paranoid just being early.
14. ADVERTISING IS NOT ABOUT BUYING, IT'S ABOUT ATTENTION
Contrary to popular belief, advertisers do not earn money directly from people who view ads. Instead, they pay platforms (such as Google, Meta, or TikTok) based on impressions, views, or predicted engagement. The platform’s role is not to force purchases, but to optimize exposure.
Even when a user skips an ad instantly, value is still created,
The system records how fast the ad was skipped,
Whether sound was on,
Whether the video was watched full-screen,
Whether the user had seen similar ads before
This behavioural trace feeds machine-learning models that improve future targeting. In other words, your non-action is still data.
Some users are adamant: they skip ads, refresh pages, restart devices, and refuse to engage. But behind every platform, they are fully aware of this group (after all some spend millions to develop the platform to make it fool-proof to this adamant group)
This "adamant" group are quickly classified as:
Low engagement,
Low conversion probability,
Ad-resistant users
Rather than fighting them, the system deprioritises them:
Advertisers bid less for their attention,
Fewer high-value ads are shown,
Generic branding or platform-owned ads replace performance ads
This is not punishment, it is optimisation. From a market perspective, not every user is worth converting.
15. BRAND ADVERTISING : EXPOSURE WITHOUT ACTION
Not all ads aim for clicks or sales. Brand advertisers pay simply for:
Recognition,
Familiarity,
Recall
A logo seen repeatedly, even for seconds, shapes future decisions. Months later, a purchase may feel instinctive rather than influenced. The value of advertising, therefore, often lies in delayed and diffuse impact, not immediate conversion.
16. THE QUIET CONCLUSION
Skipping ads does not break the system. It merely shifts how the system values you
The algorithm does not argue. It observes, categorises, and reallocates.
The deeper question is not whether ads are annoying, but whether human behaviour should be continuously harvested, inferred, and monetised without genuine understanding or choice.
That question is no longer theoretical. It is already shaping regulation, AI governance, and the future of the internet itself.
And perhaps the most telling irony of all:
Those who resist ads most fiercely still help train the very systems they distrust, simply by being observed.
