The numbers that look like reach are most likely, overwhelmingly, mostly noise.

That is not hyperbole. In a recent Heroik whitepaper, we went under the hood of one of the most important numbers in modern marketing: the impression.

We looked at how major digital platforms define and count impressions, how those definitions compare with what a normal person assumes the word "impression" means, and what remains when you begin removing the things that never resulted in meaningful communication with a human being.

The answer should make anyone responsible for a marketing budget pray they wore their brown pants today. 

When you account for bots and invalid traffic, whether the media ever appeared where a human could actually see it, whether the person paid enough attention for the message to have a realistic chance of triggering recall, and then whether that human was even relevant to the organization, the genuine audience can fall below 1% of the number being reported.

Less than 1%. Read that again. 

Now remember what we have built on top of that number.

Reach. CPM. Advertising efficiency. Audience size. Channel comparisons. Media plans. Campaign forecasts. Brand-awareness claims. Budget allocations.

Entire strategies. Billions upon billions of dollars in digital spending. And we are using an enormously inflated number at the foundation of the math.

That raises a simple question:

Why are you organizing your entire digital strategy around sawdust and noise instead of signal?

What Does An "Impression" Actually Mean?

There is a language problem hiding underneath the measurement problem.

When an ordinary person hears the word impression, they assume something was impressed upon somebody - specifically, their mind, memory, attention.

A person saw it. It registered to their consciousness. Maybe they remembered it. At minimum, some meaningful contact occurred between a message and a human mind.

Surprisingly, that is NOT necessarily what the digital advertising industry means for impression. In fact the industry standard definitions, are not even met in many cases by the 3 big companies with the largest market share of the ad business. 

On major platforms, the technical event required to count an impression can be extraordinarily weak.

Google's display-ad counting can begin with the advertisement downloading to a device. Meta has documented impression-counting conditions on portions of its network where greater than zero pixels appearing for greater than zero seconds can count. Feel free to look those up. I’ll wait. 

Think about that. The media “loaded”. For longer than zero seconds.

Impression.

No requirement that you thoughtfully looked at it. Fine. But no requirement that you consciously processed it either, let alone saw it. 

No requirement that you had a scientific chance of remembering it - meaning saw it long enough to have a shot at cognitive recall. 

In some circumstances, the technical event can occur before the thing that most people imagine when somebody tells them, "Your advertisement received an impression." Loaded on a device? No one saw it.

Meanwhile, independent research on attention and memory tells a very different story. The AIR research uses five seconds of active attention as its minimum floor because that is the lowest tested exposure duration for which the underlying research found evidence of memory encoding. Even that is a floor, not a guarantee that somebody will remember your message.

So we have two radically different concepts hiding behind one word.

Claimed impression: a “technical event” occurred with a device and/or server.

Actual human impression: a human being received enough of the message, with enough attention, that it had a plausible opportunity to possibly make an impression on the brain.

Those are not remotely the same thing. Yet organizations routinely treat them as interchangeable. The entire industry is propped up on this wordplay. 

The Human Internet Is Smaller Than The Dashboard Suggests

Before we even get to attention, we have another problem.

A surprising amount of Internet activity is not human at all.

Cloudflare reported in July 2026 that more than half of Internet traffic is now non-human, driven by automated agents and bots. That statistic describes Internet traffic broadly, not advertising impressions specifically, and the distinction matters. But it demonstrates the scale of the problem: automated activity is no longer some tiny contaminant sitting around the edges of the human Internet. It is a major part of the Internet.

Now ask yourself what numbers you are looking at every week.

Website traffic. Views. Impressions. Engagement. Followers. Reach.

And then ask the uncomfortable question:

How much of this represents actual people?

Because you are probably using those numbers to make decisions.

Your executive team is using them. Your marketing agency is using them. Your social team is using them.

Someone is calculating a return against them. Someone deciding what channel deserves another $100,000 because of them. Someone is declaring another channel "too expensive" because it cannot generate numbers that look remotely as large.

But huge numbers are remarkably easy to produce when the definition of success includes machines, invisible media, momentary rendering events, irrelevant people, and humans who never consciously processed the message.

Less Than 1% Changes The Conversation

Heroik's AIR research breaks the problem into stages.

First, we distinguish the number claimed by the platform from what we call a Likely Real Impression, or LRI.

LRI asks a much harder question:

Did the communication actually reach a human and receive enough active attention that it could plausibly trigger the brain's recall function?

To answer it, the model in our research worked through several compounding forms of attrition, including invalid and bot traffic, placements the user never reached, background or inactive tabs, abandonment before creative rendering, banner blindness, whether the person ever fixated on the media, and whether that attention reached the minimum duration used by the model.

The base case lands around 0.5 Likely Real Impressions for every 100 claimed impressions.

About 5 per 1,000. Yeah… it’s THAT bad. 

In other words, the reported number can be roughly 200 times larger than the number of communication events that clear even that human-attention threshold.

And then comes the part that makes the number smaller again.

Of The Real Humans, How Many Actually Matter?

Getting a human being to genuinely notice something still does not mean you reached your audience.

Who was that human? Was it a prospective buyer? A donor? A decision maker? A member of the community you actually need to reach?

Or was it:

• a competitor studying your positioning
• an existing customer
• a vendor
• a researcher
• a student
• someone outside your geography
• somebody outside your buying profile
• somebody who likes the subject but could never purchase
• somebody years away from any decision
• somebody who simply has no relevance to your mission

That distinction matters enormously.

The AIR framework therefore adds another filter after LRI: the Qualified Impression.

An LRI says:

A human probably received this.

A Qualified Impression asks:

Was it a relevant human?

The result is smaller still.

That is how we arrive at the brutal conclusion underneath the research:

When you move from what the system reports toward the human beings who actually received the communication, and then toward the subset of those humans who were relevant, you can be dealing with less than 1% of the apparent audience.

So take the big impression number on the screen.

Now mentally move the decimal point. Then keep going.

Then redo your economics. That is the exercise.

Sawdust Makes Things Look Bigger

This is what I mean by sawdust. Sawdust has been used as filler precisely because it creates volume cheaply. Stuff something with enough sawdust and it looks substantial. Build on enough of it and you can create the appearance of mass without much underneath it.

That is a better metaphor for what has happened to digital measurement.

We have spent roughly two decades stuffing digital economics with progressively larger numbers.

An impression sits underneath reach. Reach helps rationalize audience value. Audience value helps rationalize advertising prices. Those prices help rationalize platforms.

Platforms help rationalize agencies, technologies, services, strategies, media plans, valuations, and enormous ecosystems built on top of them.

Sawdust on top of sawdust. Noise on top of noise. Fake tits masquerading as all natural. 

Layer upon layer of activity whose apparent scale becomes evidence for the next layer.

Eventually you get an enormous structure. The problem is what the structure is standing on.

If the foundational number exaggerates meaningful human communication by one or two orders of magnitude, everything calculated from that denominator deserves another look.

Not because every digital channel is useless.

Because your perception of what you bought may be radically different from what you actually bought.

Social Media Has The Same Optical Illusion

This is not only an advertising problem. Look at the way we evaluate social media.

One person has 10,000 followers.

Another has 100,000.

The immediate assumption is that the second person commands ten times the audience.

Maybe.

But how many are active human beings? (And those forced logins or cross-posts that pull you from Instagram or Facebook to threads and Reels DO NOT count). 

How many actually see the person's posts?

How many pay attention? How many are competitors? How many are vendors? How many are aspiring peers? How many are automated accounts? How many are merely part of the giant cloud of accounts that makes the platform look larger?

And of the genuine human audience that remains, how many are actually relevant enough to create value?

Some platforms deserve credit for at least confronting portions of the problem. X, for example, has publicly documented efforts against platform manipulation, spam, fake accounts, and prohibited automation, and says it has implemented additional proactive measures since 2024 aimed at removing violative accounts.

Good.

Every platform should want cleaner numbers.

But there is a structural tension here that should be obvious.

Social-platform economics benefit from scale - or even easier, the appearance of it.

More users. More activity. More impressions. More engagement. More inventory. More apparent reach.

Those numbers become part of the story told to advertisers, users, markets, boards, and investors.

So you should be extremely cautious about allowing the system that benefits from reporting enormous activity numbers to define what those numbers mean to your business.

The number may be technically correct according to the platform's definition.

The definition may still be useless for your decision.

A small audience containing 150 relevant executives who actually know who you are can be worth dramatically more than an apparent audience of 150,000 accounts generating enormous quantities of low-value activity.

That is not anti-digital. That is basic arithmetic once you stop counting sawdust as lumber.

We have to start with the real, raw, and relevant truth if we want to improve. And this truth is about to bring some major formidable work to the forefront. 

Digital's Fake Cheapness Distorts Everything Else

This is where the measurement problem turns into a strategy problem.

Digital usually looks extraordinarily cheap. 

Five dollars per thousand impressions? Wonderful.

Now compare that against:

A conference.

A trade show.

A private dinner.

A direct-mail campaign.

A relationship-development program.

A salesperson spending an afternoon with several potential customers.

A small executive roundtable.

A face-to-face conversation.

Those channels look expensive because we tend to count them using completely different denominators.

Put a $5 CPM next to the cost of getting a qualified executive into a room and digital appears impossibly efficient.

Except one number may mean:

A single pixel loaded on a device, and registered 1,000 times.

The other may mean:

We had an uninterrupted 20-minute conversation with a human being who can actually buy.

Why are those being compared as though they measure the same commodity?

They don't.

Heroik's AIR research modeled this problem using a B2B service example.

At the headline unit-cost level, digital display looked overwhelmingly less expensive.

Then we kept going.

The modeled $20,000 digital display investment produced three transactions, giving it a customer acquisition cost of roughly $6,667.

The modeled $15,000 trade-show investment produced 30 transactions, giving it a customer acquisition cost of roughly $500.

The trade show looked catastrophically expensive using the wrong denominator.

It looked remarkably efficient when the denominator became customers.

Same money. Same channels. Same outcomes. Different measurement.

Opposite strategic conclusion. That is the danger.

Bad Digital Math Has Penalized Face-To-Face Communication

For years, we have been told that digital scales and human interaction does not.

Technically, that is true. One piece of media can theoretically be distributed to millions of devices. One human conversation can only happen among the people in the conversation. But distribution alone is not communication. Anyone can litter flyers in a parking lot.

Communication is not attention. Attention is not relevance. And relevance is not trust.

A room containing twelve of the right people can be more economically valuable than a campaign allegedly reaching twelve million of the wrong ones.

That becomes particularly important in complex B2B environments.

People need context. They need familiarity. They need repeated exposure. They need confidence. They need to believe you understand their world. They need trust.

A useful face-to-face conversation begins with several enormously valuable facts that digital media often struggles to prove.

There is definitely a human. The human is actually present. The human is actually paying some degree of attention. 

You usually know who the human is. You can establish whether they are relevant. You can adapt the communication in real time.

And if they gave you twenty minutes, they gave you twenty actual minutes. You do not need an attribution model to determine whether their browser tab was open.

That is why small rooms remain so powerful. Not because the Internet stopped mattering. Because attention density matters.

Put ten relevant people around a table and you may have ten incredibly high-value communication opportunities.

Blast something toward 100,000 accounts and congratulate yourself on the number, and you may simply have purchased an industrial quantity of sawdust.

And you may be thinking - we should just spend more to reach more. That’s fine. So long as you’re prepared to spend about 100X more to get what you think you’re currently getting. How do those numbers look to you?

A Disclosure From A Digital Operator

There is something we should disclose here. Heroik is a digital operator. We build digital systems. We deal with these platforms. We advise organizations on digital strategy. In other words, we have no financial incentive to tell you that the foundational math underneath large portions of the digital industry deserves to be challenged.

It would be much easier to leave the mythology intact.

More impressions. More reach. More traffic. Bigger audience.

Look how great everything is.

Send the next invoice.

But that is exactly the kind of bullshit we do not want to build a business on.

An enormous industry has spent the better part of two decades teaching organizations to expect digital economics that often do not survive contact with actual human behavior.

Then everyone wonders why there are so many business owners, executives, and marketing leaders who have spent significant money on digital initiatives and feel like the return never matched the promise.

Maybe the customers are not all stupid. Maybe the problem is not always that the campaign needed another quarter.

Maybe the market was trained to expect an ROI using numbers that dramatically exaggerated how many meaningful human communication events were occurring in the first place.

If you tell somebody they reached 100,000 people, they are going to develop expectations based on reaching 100,000 people.

Imagine a startup running a campaign to validate their idea operating on that expectation.

If the number of humans who meaningfully received the communication was closer to hundreds than hundreds of thousands, that changes what a rational person should expect.

And if the number of relevant humans was smaller still, it changes it again.

That is not a small reporting discrepancy.

That screws up the entire cost structure used to build the strategy.

Digital Still Matters. It Just Needs An Honest Job Description.

None of this means digital marketing does not work. We are digital operators. Obviously we do not believe that. Digital is extraordinarily useful.

It creates discoverability. It reinforces familiarity. It distributes knowledge. It gives prospects somewhere to investigate you. It supports sales conversations. It allows retargeting. It keeps an organization present between direct interactions. It enables sales. 

It can create signal at enormous geographic scale. It can support trust built somewhere else.

And in the right categories, with the right economics, it can absolutely produce customers.

But digital should have to earn its place in the strategy using the same honest economics as everything else.

It should not receive strategic privilege because its platforms can generate the largest appearing numbers.

So stop asking only:

Which channel gives us the cheapest impressions?

Ask:

What did we actually buy?

Did a human actually receive it?

Did it receive enough attention to have a plausible chance of registering?

Was it the right human?

Did that communication create a meaningful next step?

Did it enter pipeline?

Did it produce a transaction?

What did the transaction actually cost us?

Now compare channels.

Suddenly the conference does not look so expensive. The private dinner starts looking interesting. The trade association starts making more sense. The referral network becomes extraordinarily valuable. The direct-mail campaign deserves another look.

The small executive roundtable may be one of the cheapest channels you have.

And your allegedly inexpensive digital reach may become considerably more expensive once you stop dividing its cost by things that never represented meaningful reach in the first place.

Precision Attached To The Wrong Unit Is Still Wrong

Digital platforms gave us extraordinary measurement precision. That was an enormous technological achievement. It also created a dangerous psychological shortcut. Because something can be counted precisely, we assume the thing being counted is meaningful.

A server can tell you with extraordinary precision that an event occurred. That does not mean a human received a communication.

A platform can report precisely 100,000 impressions. That does not mean 100,000 people saw you. It certainly does not mean 100,000 relevant people saw you. And it definitely does not mean you should compare the economics of those 100,000 technical events against 100 face-to-face conversations as though both numbers represent the same thing.

Precision attached to the wrong unit is still wrong.

The mission was never to accumulate impressions.

The mission is to earn attention. Create influence. Remain relevant. Build trust. Produce the human decisions that actually move an organization forward. Enable the sale. 

That is the Fight for AIR.

Attention. Influence. Relevance.

And if the metrics governing your digital strategy cannot distinguish those things from the sawdust surrounding them, stop asking how far your message supposedly traveled.

Ask the harder question:

Who actually received it?

The Deeper AIR Research

This article only surfaces part of the work.

Heroik has developed a deeper AIR Standard whitepaper examining the measurement problem from the ground up, including how major platforms define impressions, the derivation of Likely Real Impressions, bot and invalid traffic, visibility and attention attrition, audience relevance, Qualified Impressions, search's Viable Buyer Rate, platform-defined "conversions" versus actual pipeline, and cross-channel economics using comparable measures such as customer acquisition cost.

The research is intentionally not published in full.

It represents working Heroik intellectual property, including the methodology used to normalize communications channels around actual human attention, relevance, pipeline, and economic outcomes.

We make the deeper AIR research available selectively to qualified leaders and organizations making serious decisions about media strategy, customer acquisition, advertising economics, communications investment, and resource allocation.

If your organization is allocating meaningful money based on impressions, reach, CPM, clicks, traffic, audience size, or platform-defined conversions, the useful question may no longer be whether your numbers look good.

It may be whether the numbers mean what you think they mean.