Nine rows, nine N/As. I read the packet twice on a Tuesday morning in Cape Town, coffee going cold, and what bothered me wasn't the emptiness. It was the formatting.
The document was a nine-dimension research framework — technical architecture, token economics, market structure, ecosystem position, regulatory exposure, team and governance, risk, narrative, industrial transmission — and every single cell returned the same three characters. Below the table sat a list of missing inputs: no title, no source link, no timestamp, no author stance, zero information points. To its credit, the analyst refused to invent. Then it did something far more dangerous: it shipped the refusal in the visual language of a finished report. Cover slide, section numbering, confidence tags. A refusal to analyze and a completed analysis look identical in a PDF.
So the question isn't whether the analyst was lazy. It's what happens inside a reader's risk model when a cell says N/A and the market is green.
Crypto research has industrialized. Frameworks are no longer private tools; they are products, sold by the slide, priced by how closely they resemble other decks that raised money. The best ones share a single discipline: every conclusion must trace back to a numbered information point in the source, marked as explicit or inferred. I believe in that discipline the way I believe in tests. In 2017 I spent four months auditing early ERC-20 implementations for three projects around Cape Town, and the only reason anyone took a young woman in that circle seriously was that I published the flaws on GitHub with reproduction steps instead of asserting them in a Telegram group. Verifiability was the argument. Education is the only true decentralized currency, and provenance is how you mint it.
The nine dimensions themselves are reasonable. They ask what an allocator should ask: what does the code actually do, who captures the fee, where does the token supply come from, which regulator has jurisdiction, who holds the upgrade key, what is the narrative versus the number. In a bull market these grids get sold as completeness. A founder hands one to a family office, the family office hands it to an investment committee, and somewhere in that chain the grid stops being a tool and becomes a receipt.
But every framework carries a shadow, and this one's shadow is geometric. A full grid and an empty grid are the same shape. When capital is abundant and diligence is outsourced three levels deep, shape is what gets purchased. That is not a new pathology — this industry has always been excellent at manufacturing the problem its next product solves — but it is new in research, where the product is supposed to be the absence of narrative.
Start with the footnote, because the footnote is where the honesty lives. N/A, the framework says, means "cannot evaluate," and must not be read as a low score. It even warns against the exact misreading I'm about to make. Except warnings don't survive contact with a spreadsheet.
In every model that sizes a position, an unknown gets one of two treatments: it is dropped, or it is defaulted. Dropping it is survivorship bias wearing a lab coat. Defaulting it — to zero, to neutral, to "probably fine" — is worse, because it hides inside an average.
There's a reason this pattern feels familiar to anyone who has written Solidity. A mapping in the EVM returns the zero value for a key that was never set. Ask for a balance that doesn't exist and the contract hands you a perfectly well-formed zero. The compiler will not warn you. The type system is satisfied. The only defense is an existence check — a second mapping, a boolean, one extra line that nobody writes because the happy path never needs it. I learned to write that line in 2017, after watching a team ship a balance lookup that could not distinguish "no stake" from "stake of zero," then pay out accordingly. The missing existence check is the most expensive line of code in this industry, precisely because it is invisible at the moment it matters.
An N/A cell is a volatility multiplier, not a stability signal. Nine N/As do not mean an asset is unrated and therefore average. They mean the variance term is unbounded, and unbounded variance does not leave expected value intact — it annihilates position size. Kelly arithmetic is unforgiving here: as uncertainty in the payoff distribution grows, the fraction of capital you can rationally commit collapses toward zero. Not because the edge disappeared. Because you can no longer prove you have one. A reader who sees nine N/As and thinks "no red flags" has done the equivalent of reading an unset mapping as a zero balance and buying more.
This is not abstract in the current cycle. Exchange distribution is decaying in plain sight — launchpad returns that once printed triple-digit multiples now barely clear double digits — and yet the research covering those launches still scores them on a nine-point grid without once asking what happens to a token when the venue's traffic advantage expires. The grid is full. The question is missing.
Which brings up the second thing the document couldn't see, because it was busy apologizing: zero information points is the most information-dense finding in the packet. A blank analysis is an analysis of the analyst. Nine N/As tell you nothing about the asset. They tell you a great deal about the pipeline — no full text captured, no source URL, no publication timestamp, no stance extracted. In audit terms, you don't grade the finding; you grade the evidence, and the chain broke at intake. That is not a neutral outcome. It is a specific, reproducible, fixable defect, and it will recur on every asset that enters the same pipe.
Which raises the economics nobody wants to discuss while everything is up. Traceability is expensive. Numbering every information point, quoting it, classifying it as explicit or inferred, attaching confidence — that is hours of labor per asset, and it produces a document that is uglier and slower than the alternative. The same cost asymmetry is quietly reshaping Europe: MiCA's headline fight is over stablecoin reserves, but what actually kills small teams is the CASP documentation burden, where a five-person project must produce the paper trail of a bank. Compliance cost scales with headcount, not with revenue. Research provenance scales identically, which is why, in a cycle where capital is free, almost nobody audits the audit. They buy the shape.
Where the framework got it right — and it did — was naming its own trap: empty analysis mistaken for completed analysis, with a hard gate that refuses to advance when information points equal zero. That is a circuit breaker, and I have deep affection for circuit breakers. In code and in research alike, the most valuable function in the file is the one that stops execution. In 2021, working with ten indigenous South African digital artists, we found that roughly 60% of secondary sales on major platforms simply didn't pay royalties. The fix wasn't a better marketplace. It was an enforcement check on every transfer, open-sourced, ugly, boring, permanent. Artists own their pixels; we just hold the keys.
Here's the uncomfortable part, and I say it as someone who has made this exact move on a stage. The refusal is also a product. "I will not analyze without provenance" reads as integrity, and often it is — but it is also the most defensible possible way to do nothing, and it carries no downside for the person refusing. The empty packet got published. Nobody was penalized. The nine-dimension grid absorbed the failure and converted it into a table.
And the deeper blind spot is that the grid maps 2017's failure modes. Supply tables, unlock cliffs, governance turnout, FDV-to-revenue — those are the questions that lost money a decade ago. The losses I sat with in 2022 came from somewhere else entirely. Signer hygiene. A single EOA holding admin rights over a treasury. Founders passing seed phrases to each other over text message during a launch window. Under Code & Conversation that winter, fifty one-on-one sessions, we audited the legacy repositories of dead projects looking for structural lessons. Almost none of them died from tokenomics. They died from an unchecked function and a human being who trusted a shape.
So: a framework that scores nine dimensions but never asks who holds the keys has itself returned an N/A on the only question that matters. Completeness is a rhetorical property, not a technical one. Any grid can be filled. Very few can be defended.
Open source is not a license; it is a promise. Research should be held to the same standard. The scarce good in this cycle will not be another nine-dimension grid with confident typography — it will be the audit trail beneath a single number, short enough to read and specific enough to check. Every line of code is a hand extended in trust, and so is every claim.
So ask yourself, honestly: what would it take for you to trust a figure you did not compute yourself? If the answer is "a well-formatted PDF," you have already priced the unknown at zero — and somewhere, a mapping you never set is quietly returning that zero back to you.