I De-Anonymized 90% of My Website Traffic 20 Years Ago. It Still Couldn't Tell Me Who Was Ready to Buy.
Twenty years ago, before third-party cookies fragmented everything and before mobile splintered attribution further, Dale W. Harrison built something most intent data vendors today still can't: a system that de-anonymized roughly 90% of website traffic down to a named individual, not just a company. He tracked everything those people did. He built sophisticated statistical models scoring them on likelihood to buy. And then, unlike every intent vendor operating today, he actually checked his work against what happened next.
The result: each of his two buckets — “likely to buy” and “not likely to buy” — was about 60% correct and 40% wrong. A small, genuine improvement over a coin flip. Not the confident, precise scoring the entire industry has been selling ever since.
That's the personal history underneath Episode 10 of B2B Effectiveness: Evidence-Based Marketing Ideas for B2B Practitioners, where Diego Sosa joins Dale again to lay out, in the most detail the series has offered yet, what should actually replace the broken MQL — not a better version of intent scoring, but a different foundation entirely.
You don't need to know why. You need to know that it works.
Diego's opening contribution reframes the whole problem through medicine. A large share of drugs on the market work without anyone fully understanding the molecular mechanism — many antidepressants and anti-anxiety medications fall into this category. Viagra, by contrast, has a well-documented mechanism. Both categories of drugs get approved and prescribed, because what regulators and doctors actually require isn't a causal story — it's proof of a reliable correlation between the intervention and the outcome, measured against a control group. Marketing rarely holds itself to even that bar.
Buyer-seller fit: the idea has two halves, not one
The replacement framework rests on a distinction the show has been building toward for weeks: it's not enough to know a buyer is willing to buy from someone like you. Your own sales organisation also has to be equipped to close them. HubSpot holds roughly 7% of the overall CRM market against Salesforce's 30% — but inside the Fortune 1000, that gap turns into 85% versus 0.4%. Large companies are not choosing the better product. They're choosing the vendor they trust will still exist in forty years, which is why a six-month-old company with a superior product still won't win Exxon's business against an established, more expensive incumbent.
Seller fit works the same way in reverse. Dale's biotech-era example: two companies selling an identical product, one selling almost exclusively to individual university researchers, the other successfully navigating the far more lucrative but far more bureaucratic procurement processes of Pfizer and Eli Lilly. Same product. Completely different sales motion, completely different skill set, completely different revenue per deal — and neither company's sales team was equipped to succeed at the other's approach.
Slot machines, not vending machines
The framing device Dale keeps returning to: stop thinking about sales like a vending machine (put in effort, get a predictable result) and start thinking about it like a portfolio of slot machines with different payout odds and different payout sizes. A machine that pays off 1% of the time but returns $1,000 on a $1 wager beats a machine paying off 10% of the time for a dime. You cannot evaluate either the win rate or the payoff size in isolation — and you cannot judge a machine by one lucky pull, because a single win tells you nothing until you've accounted for the full cost of every loss it took to get there.
This is also where two of the most common optimisation targets in B2B marketing get directly challenged. Chasing a higher win rate in isolation is a trap: a 50% close rate on trivial deals can be worth far less than a 5% close rate on deals large enough to make up for the other 95%. Chasing a shorter sales cycle is largely bogus for the same reason — sell-cycle length is mostly a property of who you're selling to, not a lever marketing pulled. The one metric that actually rolls both buyer fit and seller fit into a single, decision-relevant number is revenue generated per unit of sales effort invested.
The data was already sitting in your CRM
Rather than paying an intent vendor for a generic, everyone-gets-the-same-score product, the model the show is building looks backwards into a company's own closed-won and closed-lost history to find what genuinely, consistently differs between the two — not what closed-won accounts have in common with each other, which is close to meaningless on its own (the now-familiar example: 100% of closed-won accounts have vowels in their name, and so does nearly everyone else). It's the direct B2B equivalent of a packaged-goods company reverse-engineering purchase data to build demographic targeting — a practice that's been standard in consumer marketing for the better part of a century, and one B2B has largely ignored while chasing intent instead.
Click monkeys, and the buyers who show up out of nowhere
Dale's original validation work also surfaced two recurring, opposite failure patterns that any scoring system has to account for. “Click monkeys” engage with everything and buy nothing — genuinely interested in the content, never in purchasing. On the other side sat people who showed up on the day they were ready to buy with no prior digital footprint at all, because they'd used the product at a previous job, or a colleague had recommended it in a hallway conversation. Years later, that second pattern got a name: dark social. Two groups, behaviourally indistinguishable from a distance, sitting on opposite ends of actual buying readiness — which is exactly why no score built purely from digital touchpoints can reliably separate them.
More on how this comes together next week — there's more detail than one episode can hold.
