For 500 Years, Doctors Bled Patients With Leeches. Your Intent Data Vendor Is Doing the Same Thing.
For roughly five hundred years, bloodletting with leeches was standard medical treatment. Even decades after physicians in the 1800s worked out that it was probably killing people, doctors kept doing it — because patients kept demanding it. They didn't feel like they'd received real treatment unless a vein had been opened. The reviews, if Trust Radius had existed in 1820, would have been glowing.
That's the analogy Dale W. Harrison reaches for to describe the intent data industry today: high renewal rates, happy customers, great reviews — and none of it proves the product actually works. Survivorship bias explains the leeches. It also explains most of B2B martech.
This week's episode of B2B Effectiveness: Evidence-Based Marketing Ideas for B2B Practitioners has a different voice in Liam Moroney's usual seat: Diego Sosa, joining Dale for a deep, methodical dismantling of the single idea the whole show keeps circling back to — data is not information — with some of the sharpest illustrations of it yet.
Why ten downloads don't mean ten times the certainty
The starting example is a familiar one: a prospect downloads two ebooks instead of one, and a standard lead scoring system calls them twice as likely to buy. Download ten, and you're apparently ten times as likely. Dale's correction is sharper than “that's not how it works” — it's that the actual information curve looks nothing like the points curve. The first download tells you almost nothing (is this a person, a bot, a wrong click?). By the eighth or tenth download from the same person on the same topic, you've probably learned nearly everything the data is capable of telling you — maybe 92% certainty. Download 100 more after that, and you're not at 920% certain, because that number is meaningless. You're at 93%, if that. The information curve flattens hard. The points curve just keeps climbing.
The name tags at a networking event
Diego and Dale's clearest shared illustration: imagine walking up to three strangers at a networking event, each wearing a name tag. The first person you already know — the name tag tells you nothing new, though it might be someone else's first piece of information about them entirely. The second person's tag says “Mary,” and underneath it are nine more tags that also say Mary. Ten data points. Are you ten times more certain her name is Mary? No — you were already about as certain as you were going to get after the first tag. The third person's tag says Mickey Mouse. Your prior knowledge — what you already understand about how the world works — tells you that's almost certainly false, no matter how many tags they're wearing. Information isn't a property of data alone. It's what happens when data meets what you already know.
Intent data's fatal, unanswered question
Dale's challenge to intent data vendors is almost insultingly simple: show one graph. Plot engagement score against actual purchase outcome over a couple of years, and if the relationship is real, it should look like a clean, rising line. He's built that graph himself, repeatedly, using real vendor data (Bombora, 6sense) against multi-year sales results at large companies. Every time, it comes back as a random cloud of dots. No line. No correlation. “Extraordinary claims require extraordinary proof,” as Diego puts it — and in years of asking, neither of them has ever seen a single vendor produce that graph, because they already know what it would show.
You can't work backwards from success
A recurring analytical mistake gets a clean medical parallel: if you studied 5,000 measles patients, you'd find 100% of them have a fever. Concluding that everyone with a fever has measles would obviously be wrong — but it's structurally identical to a claim the show keeps encountering, that some percentage of closed-won accounts visited the company's LinkedIn page, therefore LinkedIn visits predict revenue. Dale and Diego's own pulled data shows the real picture: roughly 20% of closed-won deals had visited the LinkedIn company page — and about 30% of closed-lost deals had too. Closed-lost accounts were more likely to show that “positive” signal, not less. A pattern only means something if it's absent from the failures, and this one wasn't.
Billboards, branded search, and the attribution shell game
Dale's out-of-home billboard experience supplies a instructive counter-case: put up a billboard, and web traffic and branded Google searches visibly spike within the hour, as people spot it, remember a keyword, and search it later. That's a real, measurable effect — attribution just can't see the billboard as the cause, so credit quietly flows to whichever channel happened to be watching. Dale connects this to a scam he's seen repeatedly in agency-run Google Ads accounts: blending high-converting branded search (people typing your company name because they already know you) into the same campaign as generic, unbranded terms, which inflates the average click-through rate and lets an agency bill for traffic that would have arrived for free anyway.
Ask the question before you look at the data
Diego's own formative lesson, from his first data analytics job: a manager who insisted on looking at the data first to see what it says. Diego's pushback then is the same argument the whole episode makes now — you have to bring a question, even a rough hypothesis, before you look, or you'll generate plausible-sounding answers with no way to tell whether they're true. The practical version of this is thinking in interventions: is the outcome different because of the thing you did, or would it have happened anyway? Comparing a channel to a much older, established one after only three months — declaring “events work, LinkedIn doesn't” — ignores that the older channel had years of accumulated brand presence the new one hasn't had time to build.
The closing line is the one Dale returns to hardest: intent data vendors aren't lying about having data. They're lying about what it is. Intent is an internal mental state nobody can observe. What's for sale is digital engagement, relabelled. “Prove it” remains the standing challenge — one graph, one line — and in years of asking, still nobody has.
