The N/A Ledger: Why Empty Diligence Fields Cost More Than Bad Numbers
Hook
Last quarter I ran my standard nine-module diligence template against 47 mid-cap protocols. Same template I have used since 2020. Nine modules: technical architecture, token economics, market structure, ecosystem position, regulatory posture, team and governance, risk matrix, narrative versus delivery, and supply-chain transmission. Six hundred and twelve fields. Two hundred and eleven came back empty.
Not empty because the analyst was lazy. Empty because there was nothing to put there. No vesting table. No sequencer uptime log. No revenue line. No governance participation rate. No audit reference that resolved to a live document.
I logged all 211. Then I did the work I should have done five years ago: I sorted them by what an empty field actually means. They do not mean the same thing. Roughly a third were good news. Roughly a third were a warning. A third were a pricing problem I had been misclassifying as a knowledge problem.
The market has an enormous vocabulary for numbers. It has almost none for their absence. In a sideways tape, that gap is the most expensive blind spot on the board.
Context
In 2017, at twenty, I hand-audited 45 ICO whitepapers. I cross-referenced every named advisor against a LinkedIn record and a publication history. I kept three projects. Two of the other 42 collapsed inside eighteen months. The lesson I took was not "do research." Everyone says that. The lesson was narrower and more useful: a shared screening template converts an unbounded research problem into a bounded accounting problem. You cannot audit infinity. You can audit a checklist.
So I built one. Nine modules, 612 fields. It survived the DeFi Summer of 2020, the Terra collapse of 2022, and the ETF dislocation of 2024. It is the same skeleton I now use to configure risk parameters inside RuleBot, the copy-trading system I launched with 500 users and trained on five years of my own P&L. The template is not a research aid. It is the risk engine. The model does not read narratives. It reads field states.
And here is the structural flaw in every template ever built, including mine: templates have a completion bias. A row with a number in it feels finished. A row with an empty cell feels unfinished. Analysts are rewarded for finishing rows. So they fill them. They fill them with a proxy, a range, an estimate pulled from a Discord message, a number from a competitor's deck, or nothing dressed up as something.
That is how a bad number becomes more dangerous than no number. A bad number is auditable and wrong. An empty field is auditable and honest.
The sideways market sharpens this. In a trending tape, missing data is obscured by price. Everything goes up, so the empty fields never get priced. You only discover the gap when the exit queue forms. Consolidation is the phase where the market has time to read the footnotes. Sideways tape is not a waiting room. It is a re-pricing of information quality. The protocols that filled their rows with fiction during the uptrend are the ones now bleeding LPs while everyone stands around asking why.
So let me be specific about what the 211 empty fields in my last screen actually were.
Core: A Taxonomy of N/A
An N/A is not a neutral value. It is a signed value with a missing sign. Assigning the sign is the entire analytical job, and almost nobody does it.
Over four screening cycles I have converged on four classes. They behave differently, they price differently, and they require different actions. Conflating them is the error.
Type I โ Structural N/A
The question has no answer because the question does not apply. The field is empty by design, and the emptiness is the answer.
Example from my own screen: "data availability cost per megabyte." The field returned empty for 31 of the 47 protocols. Not because the teams were hiding something. Because the vast majority of rollups do not produce enough data for the question to have a meaningful answer. A rollup posting a few kilobytes per block against a general-purpose DA layer is not optimizing DA cost. It is running a rounding error through a billing system.
That is a Type I N/A. The correct reading is not "unknown." The correct reading is "this module does not govern this system's economics." The correct reading is also: the field is empty because the protocol does not need the thing the market is charging a premium for.
Here is where the mispricing lives. Scoring models penalize Type I empties identically to adversarial empties. A pipeline that awards points for answered fields will systematically downgrade the protocol whose answers are structurally unnecessary. That is not a market inefficiency in the abstract. It is a specific, repeatable discount applied to a specific class of asset, every cycle, by every fund running a checklist.
I have made more money from Type I reclassification than from any directional call I have ever placed. The trade is simple. Find a protocol being marked down for an unanswered question. Prove the question does not apply. Wait for a reviewer with a less rigid framework to notice.
The same class shows up in lending. Every lending protocol on the market publishes a utilization curve. Every one of them returns N/A on the field "market-clearing interest rate." That field is structurally empty across the entire sector, because the rate is set by a governance parameter, not discovered by supply and demand. The curve is a policy document wearing the costume of a price. Nobody scores them down for it. But the field is empty, and it has been empty for eight years, and the industry has relabeled that emptiness as a feature.
Type I N/As are the cheapest information in this market. They are the empties you are supposed to buy.
Type II โ Adversarial N/A
The answer exists and is being withheld. The field is empty because disclosure is against the interest of the party who holds the data.
This is the class that kills portfolios. My screen flagged 64 Type II empties out of 211. They clustered. They were not distributed randomly across modules. They concentrated in exactly three places: token unlocks, governance participation, and auditor identity.
An unlock table is the easiest document in this industry to publish. It is a spreadsheet. Its absence is never accidental. When a protocol publishes a pie chart showing 40% allocated to "Community" and no multisig signer list, you are not looking at a data gap. You are looking at a disclosure decision made by someone who knows the answer and has weighed the cost of revealing it.
Governance participation is the same. A protocol with a functional DAO publishes turnout. A protocol that routes every decision through a three-of-five multisig controlled by the founding team publishes turnout as a number that rounds to zero, so it publishes nothing. The empty field is the governance model. Read it as such.
Here is the quantitative problem with Type II empties, and this is the calculation that reorganized my entire process:
If you carry 28 unresolved binary assumptions into a position, and each one is 90% likely to be correct, the probability that all 28 are correct is 5.2%. Run the same arithmetic on 12 assumptions and you get 28%. Run it on 40 and you are below one percent.
0.9^28 = 0.052. 0.9^12 = 0.282. That is it. That is the whole trap. Individual assumptions at 90% confidence feel like prudence. Twenty-eight of them chained together is a coin flip with a bad haircut.
My 47-protocol screen produced an average of 4.5 Type II empties per protocol. Not catastrophic on its own. But the fields compound. Unlock schedule, signer identity, auditor identity, and governance turnout are not independent variables โ they are four views of the same undisclosed document. If you cannot get the vesting table, you will not get the signer list. The correlation is close to one. So the effective uncertainty is not 4.5 assumptions. It is one assumption with four names, and it is unresolved.
That is why I now treat Type II empties as a single binary, not a set. The binary is: will this team disclose under pressure? If the answer is no, every downstream field inherits the no.
Type III โ Analytical N/A
The answer exists. It is public. It is not in your spreadsheet.
On-chain data is permissionless. Extraction is not free. This is the class of empty field that most analysts misread as Type II. It is not adversarial. It is a cost problem, and it is the only class of N/A that money can reliably solve.
My extraction budget for a full screen runs 3 to 6 hours per protocol for Type III fields โ LP concentration, wallet clustering, oracle deviation history, liquidation cascade simulation. At 47 protocols that is 140 to 280 hours. I do not have 280 hours. Nobody does. So the real question is never "can I get this data." The question is: what is the expected value of the field, divided by the hours required to fill it, minus the confidence decay of stale data?
Filling an LP concentration field takes 4 hours and changes position sizing by a measurable amount. Filling a social sentiment field takes 6 hours and changes nothing I am willing to trade on. Same template, same empty cell, wildly different marginal value.
I built this into RuleBot as a hard rule. The engine does not act on any position where the Type III fill rate is below 70% within the screening window. Not because unfilled fields are dangerous. Because unfilled fields are unpriced, and an unpriced variable inside a leveraged position is a variance term with no cap.
Volatility is the tax on unverified assumptions. You pay it at the worst possible moment, which is the moment you need to exit.
Type IV โ Latent N/A
The field is empty because the thing does not exist yet. No mainnet, no revenue, no DAU, no TVL, no fee switch. The question is not unanswerable. It is premature.
This class is where the market does the most damage to itself, because latent empties get priced as optionality rather than as scheduling risk. A pre-launch protocol with 40 latent empties is not a protocol with 40 unknowns. It is a protocol with 40 future dates. Each one either arrives or it does not. That is an option chain, not an equity.
The error is treating a Type IV N/A as a Type II N/A and applying the adversarial discount. That is backwards. Latent empties deserve a time discount, not a truth discount. They should be priced off the delivery calendar, not off suspicion.
I refuse to underwrite Type IV empties on narrative. I will underwrite them on schedule: what is the next verifiable delivery date, what is the cost of the position if that date slips two quarters, and what is my hard exit if it slips four. That is it. Harvest when the soil is rich, not when it is wet.
What the P&L Said
When I retrained RuleBot on five years of my own trading history, I did not expect the model to find anything interesting in a field count. I expected it to weight the fundamentals.
It did not. Across roughly 340 screened positions between 2021 and 2025, the single strongest negative predictor of my realized return was not leverage, not sector, not entry timing. It was the count of unresolved N/A fields at the moment of entry.
Positions entered with four or fewer unresolved fields clustered in one return band. Positions entered with eight or more clustered in another. The gap between those bands was wide enough that I re-ran the classification three separate times looking for a coding error.
There was no coding error. There was a behavioral pattern I had been hiding from myself: I was entering high-N/A positions when conviction was high. The conviction was generated by the narrative. The N/A count was generated by the document. The narrative won, because I let it.
My own model had to tell me that I had been laundering missing data into confidence for years. It was not a knowledge failure. It was a discipline failure with a spreadsheet on top.
I will disclose the obvious bias: this is my own P&L, self-selected, no control group. It is not a study. It is a mirror. But the direction was unambiguous, and the direction is the part that generalizes.
Contrarian: The Real Risk Is Not Opacity
The consensus position is that opacity is the enemy and transparency is the cure. I think the consensus is off by one word.
The dangerous protocol is not the one with empty fields. Anyone can see empty fields. The dangerous protocol is the one with filled fields that are wrong, because a completed row terminates the analysis.
The incentive structure is the proof. If your diligence score rewards answered fields, and the reviewer cannot tell a verified number from a substituted one, then every rational project will generate answers. Not disclose them. Generate them. There is no penalty for a fabricated unlock schedule until forty-eight hours before the cliff.
And this is where I have to say something uncomfortable about my own industry. I run an AI system. I trained it on my own history. I know exactly how these models behave, because I have watched mine do it. A large language model asked to complete a diligence template will complete it. That is what it is optimized to do. Helpfulness is the objective function. An empty field is a failure state to a model that has been trained to be useful. So the model fills the cell โ with a plausible range, a competitor's figure, a confidently worded estimate with no citation.
Every fund that pipes diligence through a generative model is now manufacturing Type II N/As and relabeling them as data. The output looks better than my hand-built spreadsheet. It is worse. My spreadsheet has visible holes. Their report has invisible ones.
The second-order effect is nastier. Once fabricated completion becomes the norm, the protocols that answer honestly โ the ones that write "not applicable, structurally, here is the proof" โ will be scored as incomplete by models that have learned to expect a number. The market will start penalizing accurate disclosure because it does not look like accurate disclosure.
I audit the exit, not the entrance. Every N/A on the exit side โ unlock schedule, redemption path, oracle failure mode, sequencer uptime, governance capture threshold โ gets weighted several times heavier than any N/A on the entrance side. Marketing gaps cost you opportunity. Exit gaps cost you principal. The two are not the same field, and any template that scores them equally is a template designed by someone who has never had to sell into a falling book.
Takeaway
The next differentiator in this market will not be who has more data. Data is abundant and increasingly synthetic. The differentiator will be who has a disciplined taxonomy for the data that is missing โ and, harder still, the discipline to leave a field empty when that is the honest answer.
Watch two things in the next two quarters. First, whether EU disclosure requirements begin to specify not just which fields must be filled, but which are permitted to be empty, with justification. That is the single most valuable regulatory innovation available right now, and it is almost nobody's priority. Second, watch which protocols start publishing negative disclosures โ the signer lists, the turnout numbers, the failed audits. The first protocol that makes disclosure of a bad number into a credibility asset will re-price its entire sector.
Ledgers don't lie. Templates do, quietly, one empty cell at a time.
Liquidity is just trust with a speed limit. Trust is what you build in the fields you leave blank on purpose.
