说服和操纵用的是同一批手法，分界不在手法上

EN Persuasion and manipulation are not two sets of techniques. They are one set of techniques sitting on two different sets of facts. "Only three slots left" presses exactly the same button whether the three slots are real or invented, which is why trying to sort them by how sleazy a line feels never works. Ask a different question instead: say out loud, to their face, what you are actually doing — does the deal survive? "I quoted high first so the next number would look cheap" dies the moment it is spoken. "I'll hold this price until the 15th, because my costs go up next month" survives being spoken. Persuasion withstands disclosure; manipulation depends on the absence of it. Then a second, harder test: if they were fully informed and had all the time in the world to think, would they still buy? If yes, you are earning a matchmaker's fee. If no, you are earning their mistake — and that money gets repaid later as a refund, a bad review, or one extra sentence every time someone asks about you. The fix is never to drop the technique, it's to give the technique something true to stand on: a real deadline instead of a resetting countdown, a real constraint ("I take three new clients a month because I deliver personally") instead of a fake stock counter, and verifiable social proof — three named users they can call — instead of "chosen by thousands". Real things don't need protecting from verification.

EN A dark pattern is interface design that gets people to choose what they wouldn't otherwise choose: the pre-ticked renewal, the buried cancel link, the extra line item at checkout, the countdown that resets on refresh. What they share is that none of them persuades — they work by friction, paving the road you want them on and gravelling the one they want. And defaults are powerful enough that you have to own one as a decision, not an option: most people never change them. The organ-donation numbers are the cleanest demonstration anyone has: in opt-in countries, effective consent runs around 4% in Denmark, 12% in Germany, 17% in the UK, 27.5% in the Netherlands — that last one after a nationwide mailing campaign — while opt-out countries sit at 86% in Sweden and 99.9%+ in France and Austria, having done nothing at all. Same continent, same human beings. So "we just set a default, users can change it" is not true as a matter of fact. This has also stopped being purely an ethics argument: the EU's Digital Services Act bans interface designs that distort user choice, and in 2023 the FTC settled with Epic Games for roughly $245 million in consumer refunds over dark patterns and billing. Four things you can audit today: how many clicks cancelling takes versus buying, every pre-ticked box (untick one and watch how much conversion was never really yours), whether your countdown survives a page refresh, and the gap between the first price shown and the last — that gap is visible in your own refund and complaint data.

EN The gross margin on that one deal is capped, arrives this quarter, and is visible. The cost is uncapped, delayed, and — crucially — never lands with your name on it: it shows up as a customer who quietly stops replying six months later, a referral you never learn you didn't get, a meeting you were never invited to. Which means self-correction fails here structurally. The gain is booked against the deal and the cost is smeared across every deal that follows, so no report will ever show you the causal link, and "you learn from getting burned" simply doesn't happen — the feedback can't get back. Worse, a customer who feels conned doesn't just stop buying; they will spend their own money to hurt you, because retaliation against unfairness is a real and strong impulse, not pettiness. So replace hindsight accounting with three numbers that can see the bill for you: which rep signed the customers who churned inside ninety days and why; refund and complaint rates broken out by rep and by acquisition channel rather than averaged into invisibility; and referral rate by customer origin, because customers who were pushed into signing produce almost none — a permanent zero in that cell means you are converting reputation into revenue. One honest limit: this arithmetic depends on buyers being able to compare notes. In genuinely one-shot, anonymous markets, cutting corners really does pay, which is why tourist-trap food is bad. Don't pick the market and then rely on willpower — pick the market.

EN Nobody wakes up deciding to con someone. Crossing the line actually looks like this: you're one deal short of quota, this customer isn't really a fit, but they want to buy and nothing in the process stops you. In that moment you are not deliberating, you are solving. So the only defence that works is written while you're cold and no specific deal is on the table — turn the line into a list, so that when the hot moment arrives your job is execution, not judgement. Three lines is enough, in a fixed form: "I don't sell to ___, even if they want to buy." What goes in the blank has to be a type of buyer you actually met last quarter and actually hesitated over; "people who don't need it" is not a line, it's a mood. Then three ways of saying no that work far better than people expect. Give the reason and an alternative — "you're below the volume where this pays off; run it on a spreadsheet for three months and call me at X" — and that deal usually comes back larger. Shrink the deal rather than dropping it: "the full thing is wrong for now, start with this piece." And when you argue it upwards, translate: don't say it's unethical, say this segment's refund and complaint rates are several times the others and the net is negative. Management may not move for ethics; it will move for negative margin — provided you have the numbers broken out.