Inbound doesn't sell content. It sells the conclusion he reaches himself.

Two layers. One is leverage: explaining costs you linearly, while writing is one fixed cost against a near-zero marginal one — what content actually replaces is your repeated labour, not advertising. The other is correction: you will systematically pick the wrong topics, because you can't get back to not knowing. That bias is the curse of knowledge, and counting frequency is how you take the choice out of your own intuition's hands.

• Hard evidence (experimental economics) · the curse of knowledge: Camerer, Loewenstein & Weber (1989, Journal of Political Economy) showed in market experiments that people holding inside information cannot discount it when predicting how uninformed others will judge — they systematically overestimate what the other side knows. Peer-reviewed, cleanly designed.
• Moderate strength · the tappers study: Newton (1990, Stanford doctoral dissertation) had people tap out the rhythm of a familiar song; tappers predicted listeners would identify about 50% of them, and the actual rate was about 2.5%. Widely cited (it's where the Heath brothers' Made to Stick takes it from), but it's a single study in an unpublished dissertation — treat it as an illustration of the intuition, not as an effect size.

Boundaries: these two say you'll pick the wrong topic, not that a question asked three times is definitely worth writing — frequency is a cheap filter, not a guarantee. There's also survivorship bias baked in: your sample is people who already found you, so every so often you have to go ask outside the list (how to ask without being misled kindly: the Mom Test piece).

A conclusion you generate yourself sticks harder and moves behaviour more than one you're handed. Two mechanisms stack. Generating it deepens the encoding — what you work out yourself, you remember. And it routes around reactance: when someone feels pushed toward a position, they push back instinctively, and that reflex doesn't fire against their own thought.

• Hard evidence · the generation effect: Slamecka & Graf (1978, JEP: Human Learning and Memory) found that words participants generated themselves were remembered significantly better than words they simply read; Bertsch et al. (2007, Memory & Cognition) confirmed the effect in a meta-analysis. Classic, and meta-analytically supported.
• Hard evidence (classic) · arguing a case moves the arguer: Janis & King (1954, Journal of Abnormal and Social Psychology) had one group improvise a spoken case for a position while another merely listened to the same content; the speakers shifted their own attitudes significantly more. This is the original source for "make him say it."
• Theoretical framework · psychological reactance: Brehm (1966) proposed that when people feel their freedom of choice is threatened, they push in the opposite direction. It's a framework plus a large downstream literature rather than a single experiment — and it's exactly what explains why your conclusion gets discounted the moment you say it.
• Brain mechanism (handle with care): Falk et al. (2010, Journal of Neuroscience) scanned people watching sunscreen messaging and found that activity in the medial prefrontal cortex (mPFC, the region that handles "this is about me") predicted their actual sunscreen use over the following week — and predicted it better than their own stated intentions did. A single study with roughly 20 participants, so not settled fact; but the hint is practical: what predicts behaviour is the moment he connects the thing to himself, not whether he verbally agreed.

Boundaries: this move is safest in categories where knowing isn't doing (it needs equipment, time, a licence, someone to carry the risk). In categories where reading it is enough to do it, you really will lose some people who would have paid — that's the honest cost of the move. The test: if your entire value is knowing those three steps, what you're selling was already fragile.

Two mechanisms. The base rate: few people are in-market at all, so most of the low conversion you're seeing is that number at work, not your copy. And memory availability: when he buys, he can only pull a couple of names out of his head, and whether yours comes out depends on whether the situation he's in matches the cue you attached to yourself.

• Hard evidence · retrieval cues decide what comes back: Tulving & Thomson's (1973, Psychological Review) encoding specificity principle showed that recall depends on whether the cues present at retrieval match the context at encoding — being memorable isn't enough, the cue has to match. That gives a directly actionable conclusion: attach content to the situation he'll be standing in, not to the name of your category.
• Hard evidence · consideration sets are small: Hauser & Wernerfelt (1990, Journal of Consumer Research) pulled together data across categories showing that the number of brands consumers actually consider is in the single digits, far below the number available.
• Hard evidence (with a clear ceiling) · mere exposure: Zajonc (1968, JPSP) found that repeated exposure alone increases liking; Bornstein's (1989, Psychological Bulletin) meta-analysis confirmed the effect is real but noted that it saturates and can reverse — returns flatten after heavy exposure, and it doesn't hold for stimuli that were disliked to begin with. Showing up helps; it isn't linear, and it certainly isn't a licence to flood.
• Weaker (an industry estimate, not an experiment) · the "95-5" figure: John Dawes of the Ehrenberg-Bass Institute argued that roughly 5% of B2B buyers are in-market at any moment. That number is reasoned backwards from average category purchase cycles — it's a modelling estimate, not a measured constant, and it varies enormously by category (annual replacement versus once every seven years are nothing alike). Use it as an order-of-magnitude reminder, not for your own arithmetic; work yours out from your own contract cycle instead.

Boundaries: this is the easiest idea here to hide behind when nothing is working. How to tell them apart: if your content has never produced a single real conversation, that isn't a timing problem, it's a door or an audience problem (next section). Timing problems look different — someone does come to you, and says "just looking for now, maybe next year."

Mechanism one: searching is an act of self-selection. His wording gives away which stage he's at, so the precision of the door doesn't depend on your judgment — it depends on the words he typed. That's the same thing as qualifying, except he did it for you this time. Mechanism two, to use and to guard against: the more fluently something reads, the more readily it is taken to be true.

• Hard evidence (single studies, broadly supported direction) · fluency gets misread as truth: Reber & Schwarz (1999, Consciousness and Cognition) found the same statement was judged "true" more often when printed in higher contrast and thus easier to read; McGlone & Tofighbakhsh (2000, Psychological Science) found rhyming aphorisms were rated as more accurate than non-rhyming versions with identical meaning. Both had modest samples, but the wider direction — processing fluency shifting judgment — has fairly broad downstream support.
• Which makes this a line, too: writing clearly is a service. But part of the trust a reader gives you comes from "easy to read," not from "correct" — so the more fluent a passage of yours is, the more you owe it a second pass over its numbers and sources. Making something easy to understand is the job; letting ease of understanding stop people from checking is taking advantage.
• Mechanism, not experiment · search terms map to buying stage: I'm recommending this on mechanism. There's no clean public experiment behind it; ad-platform conversion data supports it, but that data is self-selected and the raw samples aren't published. Testing it yourself beats any industry report: ask your last ten closed customers how they originally found you.

Boundaries and uncertainty: the share of result pages that answer directly has been rising, which makes "rank first, then get clicked" shakier than it was — that's my read of the current product landscape, not a research finding, so don't take it as fact. Only one inference is safe: the more anyone can generate a piece of content, the less it's worth, and the more first-hand information is yours alone, the more likely it is to be cited.