Decision Accounting: A Textbook for Management Science

Organizations record money, assets, processes, employees, purchases, and sales, but not fate-setting decisions

Decision Accounting names the missing system of record for decisions such as entering a market, accepting a risk rating, proceeding despite safety concerns, or setting a BSA/AML monitoring threshold at 6% rather than 5%.

• The paper defines DA as systematic documentation of organizational decisions using standardized fields for who held authority, what alternatives were considered, what evidence was available
• SVB is the opening case: on November 15, 2022, the board received its fourteenth consecutive quarterly report showing held-to-maturity bonds at 57% of total assets
• Three months later SVB failed with 212 billion in assets and 175 billion in deposits; the FDIC post-mortem was 118 pages and used the phrase "no decision was made" twice.

The paper’s central failure mode is no contemporaneous record of why a decision happened

The Reconstruction Gap is the inability to answer why a decision was made from a record created at the time. The paper argues that this gap survives policies, audits, risk committees, and clean audit opinions.

• Lehman Brothers filed for bankruptcy on September 15, 2008; the Valukas Report took 18 months and ran 2,209 pages because Repo 105 reasoning, risk-limit rationale
• The 15-case sample spans four decades, three continents, and nine industries; every case had formal governance infrastructure
• The paper’s total for the 15 cases is more than 67.2 billion in regulatory fines, more than 978.5 billion in destroyed value, and more than 7,658 deaths.

The same missing fields appear in banking, aviation, energy, technology, automotive, nuclear, and chemicals

The table’s cases vary in proximate cause, but the paper treats the common failure as field-level absence: no reliable record of who decided, why, on what evidence, with what predicted outcome, and with what review trigger.

• Bhopal: safety review non-decision; 0.47 billion dollars in regulatory fines, 4.0 billion dollars in destroyed value, and more than 3,787 deaths.
• Deepwater Horizon: negative pressure test reasoning; 20.8 billion dollars in fines, 65.0 billion dollars in destroyed value, and 11 deaths.
• Volkswagen Dieselgate: defeat device authorization; 33.3 billion dollars in fines, 78.0 billion dollars in destroyed value, and about 1,200 excess NOx deaths.

Sixteen regimes from four continents converge on the same sixteen decision fields

The paper codes decision-documentation requirements implied by 16 regulatory and governance regimes, including the Monaco Memorandum, SM&CR, DORA, the EU AI Act, APRA CPS 230, MiCA, Basel III BCBS 239, FDA 21 CFR Part 820, FAA Part 25, SOX 404, COSO, ISO 31000, and the WHO Surgical Safety Checklist.

• The 17 fields are WHO, WHAT, WHEN, WHERE, WHY, EVIDENCE, AUTHORITY, TRAINING, REVIEW, STAKEHOLDERS, CONSEQUENCES, CONSTRAINTS, UNCERTAINTY, COMMUNICATION, PREDICTION, and SYSTEM WELFARE.
• The paper reports statistical independence for the convergence across 16 regimes at P < 0.001 using a hypergeometric distribution.
• The claim is that regulators independently arrived at the same operational need: accountability requires knowing why a decision-maker decided what they decided.

Across 75 public governance failures, WHY is empty 89% of the time and PREDICTION is empty 92% of the time

The paper’s broader evidence base uses 75 governance failure cases selected as the three most extensively documented failures in each of 25 industry categories, using public regulatory records only.

• The 89% Rationale-field absence rate and 92% Prediction-field absence rate are calculated across the full 75-case set, not the 15-case teaching sample.
• In the 15-case sample, the paper reports a 100% absence rate for WHY, PREDICTION, and SYSTEM WELFARE, but says that sample is not statistically representative.
• The absent fields most often tied to undefendable liability are Rationale, Prediction, and System Welfare.

A governance system fails if it cannot answer why within five minutes from a contemporaneous record

The Five-Minute Test is the paper’s practical benchmark for boards, examiners, judges, and prosecutors: answer from the record, not from reconstruction after the damage has occurred.

• The required answer is not an inference, memory, meeting-minute summary, or later post-mortem; it must be a contemporaneous record of the decision’s reasoning.
• The Monaco Memorandum is presented as one regulatory driver: DOJ charging decisions now consider whether compliance programs were adequately resourced and able to function effectively.
• The Caremark duty, strengthened in Marchand v. Barnhill in 2019, is presented as the liability channel for directors who fail to implement adequate monitoring systems.

The Decision Record turns regulatory questions into fields managers can complete before the crisis

The DA Framework organizes the same 17 fields into a usable record that captures identity, reasoning, qualification, context, and system impact.

• Core identity: WHO, WHAT, WHEN, WHERE.
• Reasoning: WHY, EVIDENCE, PREDICTION.
• Qualification: AUTHORITY, TRAINING, REVIEW.
• Context: STAKEHOLDERS, CONSEQUENCES, CONSTRAINTS, UNCERTAINTY, COMMUNICATION.
• System impact: SYSTEM WELFARE.

The Boeing case shows why Prediction and Review are not paperwork details

The paper says the MCAS decision chain was reconstructed 18 months after the second crash from fragments: emails, meeting notes, a design document with a single paragraph, and a test result with no attached analysis.

• The crashes killed 346 people.
• The reconstruction required a team of NTSB investigators four months.
• The paper argues that a Prediction field recording the expected outcome and timing would have triggered REVIEW within 90 days.

SVB, Fukushima, and Equifax failed through inaction that no record forced anyone to own

The paper treats non-decisions as documentable events: when a decision was required but no formal decision occurred, the WHY field becomes why inaction continued and REVIEW forces re-evaluation.

• SVB: the board repeatedly received reports showing bond concentration, but no signed record attested to approval of that risk.
• Fukushima Daiichi: the table identifies the primary gap as the seawall height non-decision, with 0.6 billion dollars in fines and more than 200.0 billion dollars in destroyed value.
• Equifax: the table identifies the gap as a compounding patch non-decision, with 0.575 billion dollars in fines and 5.4 billion dollars in destroyed value.

The Missing System Theorem explains why firms underinvest in decision records

The paper frames the problem as economic, not mainly compliance-based: private actors often bear the cost of documenting decisions, while shareholders, regulators, customers, and the public bear much of the cost when undocumented decisions fail.

• Across 61 industries, the paper reports β_W, welfare cost per dollar of industry revenue, ranging from 0.76 to 51.00.
• The industries named at the high end are firearms, cybercrime, human trafficking, and weapons proliferation.
• The shared structural feature is that decision-making processes generating harm lack a system of record.

DA records reasoning; it does not make judgment correct or detect deliberate lies

The paper limits DA to good-faith actors working under ambiguity and incomplete information. Its value is making negligence visible and negligent governance hard to defend.

• The WHY field asks decision-makers to state why a course of action was chosen; it does not prove the reasoning was good.
• The paper separates DA from fraud detection: Theranos and Wirecard are named as cases where false documentation, not absent documentation, was the core problem.
• Audit, forensic accounting, and professional skepticism remain responsible for detecting lies.

The paper argues DA could become the fourth pillar after Word, Excel, and PowerPoint

The paper positions DA as organizational infrastructure: Word describes decisions, Excel models them, PowerPoint announces them, but no standard system records the decision itself.

• is described as specifying the data model, API architecture, integration points with existing GRC platforms, and privilege architecture.
• The analogy is double-entry bookkeeping: it does not prevent bad investments, but it makes loss, misclassification, and deterioration visible and traceable.
• treats the 75-case evidence set as a starting dataset for a ten-year research agenda, not as a finished empirical base.