The Human Data Illusion

This article is based on my original LinkedIn article, The Human Data Illusion: Why We See People in Leads, and Why It Costs Millions, with a few extra thoughts added since.

Everyone knows bad data exists.

Everyone in marketing knows fraud exists.

Everyone running lead forms knows, at least in theory, that data should be verified before it touches a CRM.

And yet most companies still let junk in.

That is not just laziness. It is not just budget pressure. It is not even just bad process. It is psychological.

The usual explanations are familiar, and partly true:

• • They do not understand the risk.
• • They value volume over quality.
• • They think verification is not important.
• • They have been lucky so far.
• • They do not want to disrupt the lead flow.

Each of those plays a part. But they are symptoms, not the root diagnosis. The deeper reason is that human-looking data triggers human assumptions, and once that happens it is very hard to undo.

Defining the Human Data Illusion

The Human Data Illusion is the cognitive bias where companies treat leads as if they are real people simply because the data looks human.

A name, email address and phone number are not proof of a person. They are just fields that resemble one. But when you see a single lead, you do not just see a record. You see a potential customer. When you see a spreadsheet full of leads, it looks like a crowd. That is why companies find junk data so hard to reject.

Just as humans see faces in clouds, plugs, rocks and shadows, companies see people in data. In lead generation, we project human reality onto records that may only be human-shaped.

The four traps that keep junk leads alive

The illusion is not a single quirk. It shows up through four well-documented biases that work together, each pushing bad data further into accepted leads.

A. The anthropomorphic trap

Humans are wired to over-detect life and intent. It was safer to mistake movement in the grass for a predator than to ignore a real predator. That bias has not disappeared. If something looks like a person, or even slightly resembles a person, we instinctively treat it as one.

In lead generation, a name and email address triggers the same pattern. A record stops feeling like data and starts feeling like a person. Reeves and Nass showed that people instinctively treat computers as social actors, which supports the wider point that we humanise things very quickly.

B. Sunk cost bias

Once a company has paid for leads, rejecting them feels like admitting the spend was wasted. Teams try to work the leads harder rather than accept that the source was bad.

The sunk cost fallacy explains why companies continue with failing investments because they have already put money, time, or internal credibility into them. In lead buying, that means defending bad batches long after the data has stopped earning its keep.

C. The illusion of progress

A CRM filling up feels like growth. Dashboards go up. Reports look busy. The funnel appears to move. But none of that matters if the sales team is calling ghosts.

A full CRM can create the emotional feeling of progress without the commercial reality of progress. It is one of the easiest illusions in modern marketing to fall for, because the numbers look right even when nothing is converting.

D. Loss aversion

Rejecting a lead feels like losing something the business already owns. Even if the lead was never real, deleting it feels like a loss.

This is why companies keep bad records, recycle old leads, and argue over rejection rates with suppliers. The data feels like an asset, even when it is a liability.

The sequence effect

These traps do not just sit side by side. They stack.

• • Anthropomorphism starts the problem: "this looks like a person."
• • Sunk cost cements it: "we paid for these, we cannot waste them."
• • The illusion of progress rewards it: "look how full the CRM is."
• • Loss aversion prevents escape: "we cannot just delete them."

That is how bad data gets in, stays in, and keeps being defended long after the evidence says it should be rejected.

The cost of the illusion

The cost of the Human Data Illusion is not abstract. It shows up in wasted budgets, regulatory action and, occasionally, in headline-grade corporate mistakes.

JPMorgan paid $175 million to acquire Frank, believing it had more than four million users. Prosecutors later said the real number was around 300,000, and founder Charlie Javice was convicted of defrauding the bank. The point is not that this is the same as normal lead buying. The point is that even sophisticated organisations can be swayed by a dataset that appears to represent real people. The database looked human enough to help justify a huge decision. That is the Human Data Illusion at boardroom scale.

In 2023, the US Federal Trade Commission ordered HomeAdvisor, later Angi Leads, to pay up to $7.2 million over deceptive and misleading tactics in selling home improvement project leads to contractors. The case is a useful reminder that regulators will eventually catch up with the gap between how leads are sold and what they actually are.

In real estate, lawsuits have alleged that some platforms sold agents low-quality or fake leads from consumers who had no real buying or selling intent. Those claims still need to be tested fully in court, but the fact they exist shows how familiar the problem has become.

Industry studies and vendor research routinely estimate huge losses from fake leads, invalid traffic and synthetic engagement across digital marketing. The exact numbers are debatable and vary by source, but the direction is not. A meaningful share of marketing spend is wasted because data is treated as real until someone proves otherwise.

Fraudsters do not need to create customers. They only need to create data that looks human.

Why CPL is especially vulnerable

Cost per lead is where the Human Data Illusion bites hardest, because the payment event is the lead record itself.

• • CPM fraud needs impressions.
• • CPC fraud needs clicks.
• • CPL fraud only needs something that looks like a person.

A plausible name, email address and phone number can be enough to trigger payment if there are no verification controls. That makes CPL attractive to fraudsters and difficult for buyers, because every rejected lead feels like a rejected person.

Without verification, CPL becomes the soft underbelly of affiliate marketing and lead generation. With strong controls, real-time email and phone checks, fraud detection, server-to-server tracking, short attribution windows and source transparency, CPL can actually become one of the best places to catch fraud. Very few programmes put that spine in place.

Breaking the illusion

Verification and fraud detection break the Human Data Illusion by shifting the burden of proof. A lead should not be treated as real because it looks human. It should be treated as unproven until it survives checks.

• • Is the email deliverable?
• • Is the phone number valid and reachable?
• • Does the person match the claimed geography?
• • Is the submission tied to a real click or source?
• • Is the attribution window reasonable?
• • Is the record duplicated or recycled?
• • Is the behaviour consistent with a real user?
• • Is there evidence of consent or genuine intent?

This does not mean rejecting everything. It means moving from assumption to evidence. One useful internal metric is the percentage of accepted leads later proven fake, duplicated, unreachable or non-contactable. You could call it the human data handicap, but the name matters less than making the cost visible to the people signing the budgets.

A new way to see leads

Until we name the problem, it remains invisible. The Human Data Illusion explains why companies that should know better keep tolerating bad data. We do not just see leads. We see people. And once we see people, we find them harder to give up.

It is better to treat leads like Schrödinger's cat. They are neither real nor fake until tested. A lead is not a person just because it has a name. It becomes real when it can be verified.

A lead is not real because it looks human. It is real when it survives verification.

FAQs

References