你最有力的那几个推荐人，正是主流买家最不想变成的人

EN Early adopters buy being ahead: they don't mind the missing pieces, they patch the gaps themselves and treat that as their own competence, and "nobody has used it yet" is the selling point rather than the objection. The early majority buys not getting into trouble: they want proof that somebody exactly like them already did this and nothing blew up. So everything you stockpiled from your first customers — it's new, it's fast, we were first — works in reverse on the second group. You haven't got worse at pitching; your evidence is the wrong category. Three questions in a first meeting tell you which side you're on: when did you last replace a tool, who signs this off, and who in your industry is doing something similar. Long list of names on the second, and a ready list on the third, means you need peers who've done it — not a better feature story. And don't file the mainstream buyer under slow or ignorant: predictability is as rational a purchase as novelty. If he seems slow, it usually means you don't yet hold the kind of evidence he buys on.

EN "A 10% chance of slipping two weeks" is a risk: it can be estimated, budgeted, written into a plan. "No idea what might go wrong" is ambiguity: it can't be priced, so the safest move is not to move. Mainstream buyers stall on the second kind almost every time, so the job isn't claiming the risk is zero — that makes you less believable, not more — it's converting every vague reassurance into a sentence with a number, a name and a way out. There's also a layer nobody says aloud: organisations don't weigh expected value, they weigh who carries the blame if it goes wrong. Lowering that one person's exposure beats doubling your stated ROI. Practically: say the three things most likely to go wrong before he asks, and offer reversibility — one department, three months, stop and take your data. He will almost certainly never use the exit clause, but without it he won't start.

EN Thirty customers scattered across twenty industries gives you thirty isolated cases; the same thirty inside one industry gives you a word-of-mouth network that spreads on its own. The difference isn't the count, it's whether these people run into each other. So the test for a beachhead is never "is the market big enough" — it's do they talk to each other: a shared conference, group, ranking or association; the same jargon for the problem; and above all, can the second company copy the first one's setup almost verbatim. If they can't, you're starting from scratch every time, which isn't nailing a niche, it's repeating one. Two more checks. Could you plausibly become the default choice among these people within a year — not the biggest vendor, the default one? And write down the two adjacent squares before you start: the same people's next problem, and the same problem's next people. If you can't name them, you've picked an island — you can take it and still be stuck on it.

EN Between the result he wants and the product you sell there are always missing pieces: migrating the old data, training, wiring into legacy systems, compliance paperwork, who to call at 2am, getting people internally to actually use it. Early adopters fill those gaps themselves and enjoy it — closing them is the edge they were buying. The early majority won't fill a single one: leave one piece unowned and the whole thing reads as "can't be done here," regardless of how good your product is. So most of the work of crossing is not product work, it's writing a name in every square — you, a partner, or the customer himself, as long as he's said so out loud and it's in the plan, because "he'll handle it" and "nobody's handling it" look identical in meeting notes. Taking every square yourself is the other failure: you become a project shop where each deal is bespoke and no deal helps the next. And note that a completed whole product only counts for one niche — which is exactly why the beachhead and the whole product are the same decision.