.png)
For roughly five hundred years, bloodletting with leeches was standard medical treatment. Even after physicians in the 1800s began to understand that it could be doing more harm than good, the practice persisted. Patients expected treatment to look like treatment: if a vein hadn't been opened, it didn't necessarily feel as though anything had been done.
The fact that people kept asking for it, and that doctors kept providing it, was not evidence that it worked.
That's the analogy Dale W. Harrison uses to describe one of the central problems in B2B marketing: a product can have satisfied customers, strong renewal rates and excellent reviews without those things proving that the underlying product delivers what it claims.
This week's episode of B2B Effectiveness: Evidence-Based Marketing Ideas for B2B Practitioners explores that problem through a conversation between Dale and Diego Sosa, who joins the series in place of Liam Moroney.
Together, they unpack one of the ideas the series has been building toward from the beginning: data and information are not the same thing.
Why ten downloads don't mean ten times the certainty
The starting example is a familiar one: prospect downloads two eBooks instead of one, and a conventional lead-scoring model treats that as twice the evidence of buying intent. Download ten, and the score continues to rise.
But the information contained in those downloads doesn't increase at the same rate.
The first download may tell you very little. Was it the right person? Was the content relevant? Was the download deliberate? As more activity accumulates, your confidence may increase, but eventually the additional activity tells you less and less that you didn't already know.
If someone downloads ten pieces of content about the same subject, the tenth download might add some information. The hundredth probably adds very little. Yet a points-based system can continue assigning points to every action as though each one were an entirely new piece of evidence.
That's the distinction between data and information. You can keep collecting data long after you've stopped learning anything particularly new from it.
The name tags at a networking event
Diego and Dale illustrate the same idea with a networking exercise.
Imagine meeting someone wearing a name tag that says Mary. You now have one piece of data suggesting her name is Mary. If she is wearing ten identical tags, you don't become ten times more certain. The additional tags haven't given you nine new pieces of information. They've simply repeated the first.
The same applies in reverse. If someone is wearing a name tag that says Mickey Mouse, your existing knowledge about the world tells you to question what you're seeing. The data doesn't exist in isolation. Its meaning depends on what you already know.
That's an important distinction for anyone working with large datasets. Information isn't simply a function of how much data you have. It comes from the relationship between new evidence and what was already known.
Intent data's unanswered question
That leads to a straightforward test for intent data.
If engagement really predicts buying, there should be a measurable relationship between the two. Plot engagement against actual purchase outcomes over a sufficiently long period and, if the relationship is strong, you would expect to see some meaningful pattern.
Dale's challenge to intent-data providers is essentially to produce that evidence. He describes repeatedly looking at this relationship using real-world vendor data and finding something much less convincing: a cloud of points rather than a clear relationship between engagement and eventual purchase.
That doesn't mean engagement data has no value. It means the claim being made about what that data represents matters. If the data shows that someone interacted with your website or content, that's what you know. Turning that observation into a claim about what someone intends to buy requires evidence of the connection between the two.
As Diego puts it, extraordinary claims require extraordinary proof. In this case, the proof should be relatively easy to demonstrate: show the relationship between the signal and the commercial outcome.
You can't work backwards from success
Another analytical problem the episode explores is the tendency to look at successful customers and work backwards from their behavior.
The medical analogy makes the flaw obvious. If you studied 5,000 people with measles, you would find that essentially all of them had a fever. That doesn't mean everyone with a fever has measles. You have studied the characteristic of the people who had the outcome without checking whether it distinguishes them from people who didn't.
The same mistake can appear in B2B analysis. Suppose 20% of closed-won accounts visited your LinkedIn company page. That might sound like evidence that LinkedIn engagement is associated with winning.
But what if 30% of closed-lost accounts visited it too?
The signal is there. It just isn't discriminating between the two outcomes.
That's why looking only at your wins isn't enough. To understand whether a characteristic tells you anything useful, you need to know how it appears among both the successes and the failures.
Billboards, branded search, and the attribution problem
The episode also looks at a different kind of measurement problem: cases where something genuinely causes an outcome, but the attribution system assigns the credit somewhere else.
Dale's experience with out-of-home advertising provides a useful example. When a billboard goes live, branded search and website traffic can increase shortly afterward as people see the message, remember the brand or keyword and search for it later.
The search is measurable. The billboard's influence often isn't.
That creates an attribution problem because the system can see the final click without necessarily seeing what prompted it. The measurable interaction receives the credit even when another channel helped create the demand.
The same issue can appear in paid search. Dale describes seeing branded search terms – where someone is already searching specifically for a company or product they know – combined with generic search terms in the same campaign. Because branded searches typically convert at a much higher rate, they can make the campaign's overall performance look stronger than the incremental contribution of the advertising actually is.
The lesson isn't that attribution is useless. It's that measurement needs to account for what would have happened without the intervention.
Ask the question before you look at the data
Diego learned a version of this early in his career as a data analyst. A manager encouraged him to look at the data first and see what it said. Diego came to see the problem with that approach: without a question or hypothesis, almost any dataset can produce a pattern that sounds plausible.
The more useful approach is to start by asking what you are trying to find out.
Did the outcome change because of the thing you did, or would it have happened anyway? That's the question behind an intervention-based approach to measurement.
It also means being careful about comparisons. A new marketing channel that has been running for three months isn't necessarily comparable with an established channel that has been building awareness and demand for years. Declaring one successful and the other ineffective without accounting for that history can turn a difference in maturity into a conclusion about performance.
The data may be accurate. The conclusion drawn from it may still be wrong.
Data is not information
That distinction sits underneath almost everything Dale and Diego discuss in the episode.
B2B marketing has never had more data available to it. The challenge is that collecting more of it doesn't automatically produce more understanding. A thousand signals can still tell you less than one well-chosen signal if none of them helps distinguish the outcome you're trying to predict.
The question, then, isn't simply “What data do we have?” It's “What information does that data actually give us?”
That distinction matters particularly when the goal is to decide which accounts are worth pursuing. Knowing that an account is engaged, or even that it is likely to buy something in your category, doesn't tell you whether it is likely to buy from you. That requires a different kind of evidence: an understanding of the relationship between the buyer and the seller, and whether your business is actually well positioned to win that account.
That's the question Dale and Liam explore in Episode 9: The Buyer-Seller Fit Model – and why understanding who is likely to buy isn't enough. The more useful question is who is likely to buy from you?
