.png)
Twenty years ago, before third-party cookies fragmented the web and mobile made attribution even harder, Dale W. Harrison built something most intent-data vendors today still struggle to deliver: a system that could de-anonymize roughly 90% of website traffic down to a named individual, rather than simply identifying the company behind the visit.
He tracked what those people did, built statistical models to score their likelihood of buying, and then did the part that matters most: he checked the predictions against what actually happened.
The results were far less precise than you might expect. His two groups – people the model considered likely to buy and those it considered unlikely to buy – were each roughly 60% correct and 40% wrong. The model was doing better than chance, but only modestly. After all that data and modeling, it could improve the odds. It couldn't tell him with certainty who was ready to buy.
That experience provides the foundation for Episode 10 of B2B Effectiveness: Evidence-Based Marketing Ideas for B2B Practitioners, where Diego Sosa joins Dale to go deeper into what the series has been building toward: what a useful alternative to conventional MQL scoring could actually look like.
You don't need to know why. You need to know that it works.
Diego opens with an analogy from medicine. Many drugs work reliably without scientists having a complete understanding of every mechanism involved. What matters for approval is evidence that the treatment produces a meaningful outcome under controlled conditions.
Marketing rarely applies the same standard to its models.
A scoring system can contain sophisticated mathematics, thousands of data points and a convincing explanation of why particular signals supposedly indicate buying intent. None of that establishes that the model actually predicts the outcome it claims to predict.
The more important question is simpler: when the model says this account is more likely to buy, does it actually turn out to be more likely to buy?
Dale's early experiment is useful precisely because he measured that.
Buyer-seller fit has two sides
The framework the episode develops also brings together an idea the series has been building toward: buyer-seller fit works in both directions.
A buyer might be willing to purchase the type of product you sell and still be unlikely to buy from your company. HubSpot, for example, holds roughly 7% of the overall CRM market compared with Salesforce's 30%, but the picture is dramatically different among Fortune 1000 companies, where Salesforce has a far stronger position.
The product itself isn't the only variable. Buyers also have preferences about the companies they are prepared to work with, particularly when the purchase carries significant financial or operational risk.
The same is true on the seller side. Dale gives the example of two companies selling essentially the same product, but with very different customer bases. One had developed expertise selling to individual university researchers; the other had learned how to navigate the complex procurement processes of large pharmaceutical companies such as Pfizer and Eli Lilly.
Their products could be comparable. Their ability to win particular customers wasn't.
That's why buyer-seller fit can't stop at asking whether an account looks like a good customer. It also has to consider whether the seller has demonstrated an ability to win customers like that one.
Sales is a portfolio of bets
Dale uses the analogy of slot machines to explain why this matters commercially. Each opportunity has both a probability of winning and a potential payoff. A deal with a high probability of closing isn't necessarily more valuable than one with a lower probability if the potential revenue is dramatically different.
The same applies to sales effort. A 50% close rate on small deals can produce less value than a 5% close rate on opportunities with substantially greater lifetime value. Looking at win rate alone misses the economics of the decision.
Sales-cycle length presents a similar problem. A shorter cycle isn't automatically better if the faster deals are less valuable, and the length of a sales cycle is often heavily influenced by the type of customer being pursued.
The more useful question is therefore what return the business gets from the sales effort it invests. That brings probability, customer value and the cost of pursuing the opportunity into the same calculation.
The data is already in your CRM
The model Dale describes doesn't begin with a generic scoring system purchased from an outside vendor. It begins with a company's own history.
Look at the accounts that became customers and the ones that didn't. Then identify the characteristics that actually distinguish those two groups.
That distinction matters. Finding something that every closed-won account has in common tells you very little if the same thing is common among everyone else. The familiar vowels example makes the point: every company name contains vowels, so the fact that 100% of your customers have vowels in their names tells you absolutely nothing about who is likely to become your next customer.
The useful patterns are the ones that separate outcomes. They tell you something about why some accounts are more likely to become customers than others.
The buyers your data can't see
Dale's earlier validation work also exposed another problem with digital scoring: some of the strongest buying signals never appear in the data.
There are the “click monkeys” — people who engage enthusiastically with content, visit pages and download resources but never become customers. Their digital behavior can look exactly like serious buying activity.
Then there are buyers who appear almost from nowhere. They may have used the product at a previous company, heard about it from a colleague or simply arrived at the point of purchase through conversations that never touched a trackable marketing channel.
Those two groups demonstrate the limits of digital behavior as a proxy for buying readiness. One can generate enormous amounts of measurable activity without buying. Another can generate almost none and still become a valuable customer.
The answer isn't to keep adding signals in the hope that enough data will eventually reveal who is ready to buy. It's to understand what the available evidence can actually tell you, test it against commercial outcomes, and then use it to make better decisions about where sales effort goes.
That raises the next question: once you have a better way of deciding where to focus, how do you know whether that sales effort is actually working?
In Episode 11, Dale takes on sales effectiveness, looking at how businesses can measure the return they're really getting from the effort they put into winning customers.
