A young man came to my office last Boishakh with the typescript of his first novel, about the widow of a fish trader in Sandwip. After seven pages I knew it would not survive its second chapter. He was watching my face for permission to be a writer. I did not give it. Fifty five years above this printing press have taught me that to grant permission a manuscript has not earned is the unkindest editorial act a publisher can perform. The people in software call this faculty "product sense" and speak of it as though it were one thing, when it is two jobs joined at the hip, and almost everyone who teaches it teaches only the easier.
On one side stands the artifact, the thing the team is going to build, for whom, by what date, with what compromises around scope and headcount. This is the manuscript itself. A sincere apprentice with three weeks of customer conversations can produce a passable version. The other side is what I would call the editorial eye. It is the faculty by which you know, in the second paragraph, whether the rest of the book will hold. The same faculty tells a product manager that an onboarding flow which tests beautifully in the office will quietly die in its third month. You cannot write it down on a card. You can act on it, and when you are wrong you must eat the loss in public and figure out which of your assumptions was lying to you.
Interview boards are stuck on the artifact half because the artifact half is gradeable. You can mark a candidate against a rubric, count the cells in their tradeoff matrix, and be at lunch by half past one. The editorial half is harder to grade, so the graders retreat to what is auditable, and the prep industry, being a faithful mirror of what is rewarded, follows. You end up with product managers who can run a framework in their sleep and still ship a feature nobody wanted, because nothing taught them to feel the wrongness of it on the morning of day one.
You publish a book and it does not sell. You sit with the failure long enough to understand which of your five judgements broke first, and why you had believed it. Then you publish another, and another. After perhaps fifty such cycles, in my own count closer to seventy, you begin to recognise the shape of a manuscript that will disappoint before the proofs come back from the press. You cannot always articulate why, and you can almost always be right anyway, which is a different kind of correctness from the one rubrics measure.
Daniel Kahneman called this faculty System 1, more or less. In Thinking, Fast and Slow he insists that System 1 is trustworthy only where the feedback loop is tight and honest. Chess masters have it, physicians in the emergency room have it, and stock pickers, by his account, mostly do not, because their feedback is noisy and late. Product work, like book publishing, sits in the middle of that spectrum, tight enough that the eye is real, noisy enough that confident people are wrong all the time.
This is the dirty secret nobody puts on a curriculum vitae. People with what your trade calls "product sense" carry, at the back of the cupboard, a stack of failed projects rather than a stack of read books. The books are scaffolding. The muscle is grown by stressing it past where it would rather go and paying attention to what tore.
Bibhutibhushan, in the closing chapters of Pather Panchali, does something no manual on writing fiction has ever taught. He moves the reader from the death of Durga to the family's departure from Nischindipur in a register so exactly weighted that the proportions, one realises on a second reading, were the entire achievement. His editor at Bichitra in the 1920s presumably saw it on the typescript before the proofs were set. No rubric would have produced that recognition. Only the years of reading bad novels and good ones, until the ear knew the difference without consulting itself.
Product sense behaves the same way. The artifact half can be taught in a course of four weeks with a polite certificate at the end. The editorial half is the part one can only live through, slowly, and most teachers of the first half do not disclose that the second is missing from their syllabus. Richard Rumelt, in Good Strategy Bad Strategy, makes a parallel observation that strategy which turns out to be fluff is mostly the strategy where the diagnostic step was skipped. Frameworks unbacked by reps produce something that sounds confident and amounts to nothing.
If you are early in your craft, the counsel from a printing press in Old Dhaka is to ship more things on smaller surfaces with shorter cycles than feels comfortable. Take the loss when the loss arrives. Write the autopsy for the version of yourself who will need it in six months, not for the organisation chart that reads it next quarter. If you are hiring, ease up on the rubric. Ask the candidate about a thing they shipped that genuinely failed, and listen, three years on, for whether the autopsy is still vivid in their mouth and whether they can tell you which specific assumption broke first. The framework they remember is the weakest signal in the room.
And if you are teaching the craft, please be honest about which half of it can be taught. The frameworks half belongs in books and in courses that are well run, and you can transfer it, in good faith, to a roomful of people in a quarter. The editorial half is bruises, and what you owe a serious student is not a syllabus but the situations in which they can earn their own. I still have not found the kind sentence to say to the young man about his second chapter. He is coming back on Thursday.