Why does research produce longer stories the longer it runs? An earnings call runs an hour, the report runs thirty pages, and the business model reads more like a novel with every draft. The longer the story, the more it asks one to believe; every extra layer of belief leaves the judgment hanging somewhere new. The problem is not whether the story is well told; it is that almost nobody asks whether all of it can be cut to three sentences.
One fundamental-research framework sets the first screen brutally hard: state a company’s core, unchanging asset in three sentences — whatever remains after short-term noise is stripped away. If three sentences suffice, the simpler the logic, the more reliable it is. If they do not — complicated story, layered segments, a bet on some future pie — pass. Complexity is not depth; it is a source of error. The framework’s own first step is to strip away short-term noise and extract that unchanging core; the three-sentence test measures exactly this step’s output.
Luckin in its early years is the textbook case screened out by this standard: burning cash to expand, no stable moat, judged a value-destroying asset. One entry; no need to walk through the financials or the share price afterward. The opposite case: CIMC — global container capacity, explainable in one sentence. A business that can be stated clearly earns the next round of research. One that cannot sits on sand, and every valuation and model built on top of it sits there too.
First objection: what three sentences cannot hold is the real substance; a complicated logic means the research went deep. This reads the direction backwards. Reciting complexity as complexity is not research; it is restatement, not analysis. The depth of research shows in peeling complexity down to simplicity — stripping the noise, stripping the narrative, until what remains is the thing that does not change. Every link in a chain that cannot be stated is a point of failure; the more links, the higher the odds the judgment collapses. Research that goes deep ends with fewer words, not more. A restatement can be memorized; a judgment cannot.
Second objection: some good businesses are genuinely layered, and a three-sentence test kills them unfairly. Layered businesses can be taken apart, one layer explained at a time; if every layer reads clearly once separated, the business passes the screen. If it still cannot be stated after being taken apart, the problem is not that the business is complex — it is that no core can be isolated, and the act of taking apart is itself the stripping. If the researcher cannot isolate it, who else is expected to? A layered business that reads clearly and one that does not are two different things; the screen stops the second kind.
Third objection, the most cutting: companies built on a future pie rally the hardest, so passing on them means missing the biggest moves. Yes, it does mean missing them. But ask one more question: can anyone actually earn an advance they do not understand? And having earned it, can they hold it? What is given up was never the stock; it is the right to understand. Money outside one’s understanding cannot be held on the way up and cannot be bought on the way down — on the account it is not return, it is luck. The framework’s own rule says so plainly: what cannot be understood, pass on, never force. Missing out is not the cost of this system; it is the design.
The pity is for the people trapped by complicated stories: one story told for three years, and the storyteller cashed out long ago.
So three sentences come first, before valuation — the sharpest valuation coordinates are only for businesses that can be stated clearly. If three sentences cannot be produced, the fault is usually not in the mouth; the business itself has no unchanging core. If it cannot be said, pass. Passing is not defeat; it is the first move of research — and outside the door, no money is lost. This screen only judges formed businesses: for an early-stage company, failing three sentences is no sin — the core has not grown out yet. The screen issues no verdict on them; it only asks one to wait.
