Prescribing before you
examine isn't efficiency

Naming a solution is an identity claim. The moment it's on the table you go from "someone looking at the problem with him" to "someone selling him something", and his evaluation frame switches with you: no longer "what's actually wrong here" but "is this price right for these features". And the real value of diagnosis isn't the information you gather — it's that the conclusion comes out of his mouth. A conclusion you generate yourself and a fact you were told do not travel the same route.

• Hard evidence · professionals interrupt far sooner than they think: Beckman & Frankel (1984, Annals of Internal Medicine) recorded 74 office visits — physicians interrupted a mean of 18 seconds into the patient's opening statement, and only 23% got to finish. Marvel et al. (1999, JAMA) re-measured 264 visits fifteen years later: still only 28% finished, mean 23.1 seconds. Singh Ospina et al. (2019, J Gen Intern Med) found a median of 11 seconds. Three samples, three decades, one direction. Discount: this is clinical, not sales — read it as evidence of the impulse, not as sales data.
• Hard evidence · saying it yourself isn't the same as being told: self-referential processing — relating information to "me" — reliably activates the medial prefrontal cortex (mPFC) and posterior cingulate and produces markedly better memory (Kelley et al., 2002, Journal of Cognitive Neuroscience; Northoff et al., 2006, review); the generation effect points the same way (Slamecka & Graf, 1978). Discount: the materials are word lists and trait judgements, not purchasing decisions — don't stretch it into "good questions close deals".
• Field observation · proposing early correlates with losing large deals: the Huthwaite team (Rackham, 1988, SPIN Selling) observed roughly 35,000 sales calls: in small transactions, leading with features costs almost nothing; in large ones, successful sellers introduced solutions notably later. Discount: observational, 1980s B2B, no randomised control, and the firm sold sales training — read it as direction, not causation.

Boundary: diagnosis is not stalling, and it isn't an intake form. When someone already knows exactly what they want and is only comparing prices (repeat purchase, commodity), more questions just annoy them. One test: if answering the question gives them nothing new, don't ask it.

Every vendor sentence arrives pre-discounted, and the discount comes from the listener's inference about your motive: they know you have a stake, so everything you say gets marked down as "you would say that". The only thing that lowers the rate is a true statement that costs you — it rules out "he says it because it pays him". So what the three tiers really sell isn't the tiers; it's that you're free to say "don't buy".

• Hard evidence · statements against self-interest are more credible: Eagly, Wood & Chaiken (1978, JPSP): when a communicator's position disconfirms what the audience expected from their background and interests, participants judged them less biased, more credible, and shifted attitudes more; when the message conformed neatly to the communicator's interests, participants attributed the position to the stake rather than to the facts. A foundational result in attributional persuasion research — solid.
• Theory plus large evidence base · once persuasion knowledge fires, the object of evaluation changes: Friestad & Wright's (1994, Journal of Consumer Research) Persuasion Knowledge Model — the moment someone recognises "this is an attempt to persuade me", they switch into coping mode, start inferring motive, and discount content accordingly. Which is why "we're perfect for you" gets weaker the harder you push it: the pushing is the recognition cue. Strength: a framework with a lot of downstream evidence, so a little softer than a single clean experiment.
• Related, with discounts · admitting a weakness helps, under conditions: two-sided messages — stating the downside as well as the upside, with a response to it — persuade better than purely positive ones, especially with knowledgeable or already-sceptical audiences. Discount: effect sizes are modest, and it depends heavily on whether the weakness you admit is actually relevant to their decision. Strategic self-deprecation about something irrelevant backfires harder than a straight pitch once it's seen through.

Boundary: this only works while you're telling the truth. Turn "I don't recommend that" into a routine — always cutting the module nobody buys anyway, to look honest — and they'll see it the second time. Worse, once you start performing honesty you have to remember your performances. And a failure in the other direction: if the thing you cut is something they genuinely need, that isn't honesty, it's malpractice.

Satisfaction is a difference, not a level. The same delivery becomes disappointment if expectation was set one notch high and becomes trust if expectation was set accurately — so every promise made on signing day is pricing the future. And the difference is asymmetric: the bad side carries systematically more weight, so one disappointment takes several satisfactions to repair.

• Hard evidence · satisfaction = delivered minus expected: Oliver's (1980, Journal of Marketing Research) expectancy-disconfirmation model — satisfaction is driven mainly by performance relative to prior expectation, not absolute performance; the foundation of consumer satisfaction research for decades, repeatedly replicated. The neural version has the same shape: midbrain dopamine neurons encode reward prediction error — firing rises when things beat expectation and dips below baseline when they fall short (Schultz, Dayan & Montague, 1997, Science). Overpromising doesn't excite the buyer; it raises the baseline you then have to clear.
• Hard evidence · bad is stronger than good: the reviews by Baumeister et al. (2001, Review of General Psychology) and Rozin & Royzman (2001): across impression formation, close relationships, word of mouth and much else, negative events systematically outweigh equally intense positive ones. So a bad sale can't be costed as "one refund".
• Industry analysis · right direction, don't copy the number: Reichheld & Sasser (1990, Harvard Business Review) put a 5% improvement in retention at 25%–85% more profit. Heavy discount: a consultancy's cross-industry calculation, method not fully published, enormous variation between industries. The direction holds; don't take the multiple into a budget.

Boundary: don't turn "willing not to sell" into a pose. Using "I'd advise against buying" as an advanced closing line, hoping they'll chase you, is manipulation — and it's dead permanently once seen. Real refusal costs something: this month's number looks bad. Which is why this is hardest for a new rep on full commission, and why it belongs in how a company sets targets rather than resting on individual conscience. The test is the one from the influence-principles piece: what still works when said out loud is persuasion.

Value created by diagnosis that doesn't land on an action of theirs becomes an asset they take for free; and a meeting with no action generates no commitment pressure, so your priority on their calendar resets to zero the next day. Asking for a small action does a second job too: it's the most honest qualifying signal available — saying "very interested" costs nothing, while pulling a colleague into a room spends their internal credit.

• Hard evidence · specifying when, where and how raises follow-through: Gollwitzer & Sheeran's (2006) meta-analysis of 94 studies found that turning an intention into an if-then implementation intention produces a medium-to-large effect on completion (d≈0.65). Which is why "let's find time next week" and "Thursday 10am, you, me and W, thirty minutes" are not the same thing.
• Field observation · an advance and a continuation are different events: Rackham (1988) splits call outcomes into an advance (the buyer commits to a specific step that moves things forward) and a continuation (warm meeting, no agreed action), and notes how many salespeople log the second as progress. Discount: observational, no randomised control — but anyone who has run a pipeline can verify this one themselves.
• Fairly hard · small commitments shift later behaviour, but don't use them as leverage: the foot-in-the-door line of work (Freedman & Fraser, 1966 and successors) shows that agreeing to a small request raises compliance with a larger one later; Burger's (1999) meta-analysis confirms the effect is real but small and condition-dependent. So read it as a signal, don't press it as a lever.

Boundary: asking for an advance is not pressure, and the difference is whether the action has independent value to them — "thirty minutes with W" gets an internal bottleneck talked through even if they never buy; "sign a letter of intent" is worth something only to you. Another failure mode: the action isn't theirs to authorise in their own organisation, and you read the stall as disinterest. What to ask then isn't "are you still interested", it's "who has to nod for this to move?"