The ROI Incantation: An Archaeological Excavation of 111 Years of Expensive Bullshit
Or: How a Diagnostic Tool Became a Protection Spell, Got Buried Under Seven Layers of Contested Philosophy, Enabled the Largest Financial Frauds in Human History, and Now Powers “AI Transformation” Decisions
Part of the “2025 Must Die!” Series
Prologue: The Spell
“What’s the ROI?”
I’ve heard this phrase approximately 47,000 times across three decades in enterprise technology. It’s spoken with the confidence of someone invoking mathematical truth. It’s deployed as the final arbiter of business decisions. It lands on the table like a judge’s gavel.
And it’s complete and utter bullshit.
Not because return on investment doesn’t matter. It does. But because the phrase “ROI” has become a *protection spell* – a magical incantation that replaces thought with ritual. When someone says “What’s the ROI?”, they’re not asking a question. They’re performing a corporate exorcism, attempting to banish uncertainty through the power of a three-letter acronym.
Here’s what nobody in that boardroom knows: That number they’re waiting for – that single, decisive figure that will justify millions in spending – is built on 111 years of contested philosophy, buried assumptions, arbitrary decisions, and at least one moment where software engineers picked the simplest option because it was easiest to code.
Let me take you on an archaeological dig through the sediment. Pack a headlamp. It gets dark down here.
Layer -3: Donaldson Brown and the Formula Nobody Uses Correctly (1914)
The Origin Story
In 1914, a young electrical engineer named F. Donaldson Brown was working at DuPont. He wasn’t an accountant. He wasn’t a finance guy. He was an engineer who understood *systems*.
DuPont had a problem: they had diversified from explosives into lacquers, plastics, and dyes. Completely different businesses with completely different economics. How do you compare them? How do you allocate capital across divisions that have nothing in common except the DuPont name?
Brown created what became known as the “DuPont Formula” – a decomposition of Return on Investment into its component parts:
ROI = Profit Margin × Asset Turnover
Or, expanded:
ROI = (Net Income / Sales) × (Sales / Total Assets)
Later extended to:
ROE = Profit Margin × Asset Turnover × Financial Leverage = (Net Income/Sales) × (Sales/Assets) × (Assets/Equity)
The Point Everyone Missed
Here’s what’s crucial: The decomposition was the entire point.
Brown didn’t create ROI to produce a single number. He created it to diagnose why returns differed between divisions:
High margin, low turnover? → That’s your specialty chemicals business
Low margin, high turnover? → That’s your commodity business
Same ROI, different paths? → Different management approaches needed
The formula was a diagnostic tool, like a doctor’s examination. The blood pressure reading isn’t the diagnosis – it’s data that informs the diagnosis.
The Cargo Cult Extraction
But something happened between 1914 and now. The decomposition got stripped away. The diagnostic components got forgotten. And “ROI” became a single number that people calculate, compare to a hurdle rate, and use to make binary decisions.
It’s as if doctors forgot about diagnosis and now just shout “BLOOD PRESSURE!” at patients and expect them to get better.
Brown built a circuit diagram for financial performance. A century later, accountants turned it into a checkbox.
Layer -2: The German Bilanztheorien Wars (1900-1960)
But Brown’s formula needs input data. Specifically, it needs costs. And this is where we descend into genuine philosophical horror.
The Fundamental Question Nobody Agrees On
In the early 20th century, German academics were engaged in fierce debates about a question that sounds simple but isn’t:
What is the TRUE cost of something?
Not “what should we charge” or “what can we recover.” What *IS* the cost, as a matter of objective reality?
This spawned decades of competing Bilanztheorien (balance sheet theories):
1. Statische Bilanztheorie (Simon, 1899)
- Question asked: What could creditors recover in liquidation?
- Balance sheet purpose: Snapshot of wealth at a point in time
- Asset valuation: Realizable values
- Philosophy: Protect creditors by being conservative
2. Dynamische Bilanztheorie (Schmalenbach, 1919)
- Question asked: How efficiently did we operate this period?
- Balance sheet purpose: Performance measurement over time
- Asset valuation: Assets are “suspended items” – future expenses/revenues
- Philosophy: Income statement primary, balance sheet secondary
3. Organische Bilanztheorie (Schmidt, 1921)
- Question asked: Can the firm replace what it consumed?
- Balance sheet purpose: Maintain real capital (Substanzerhaltung)
- Asset valuation: Replacement values (Tageswerte), not historical cost
- Philosophy: Recognize inflation effects explicitly
4. Rieger’s Nominalistic Theory (1927)
- Question asked: What profit accrued to capital owners?
- Balance sheet purpose: Only liquidation provides “true” settlement
- Philosophy: Everything before liquidation is provisional
The Multiple Methodenstreite
The Germans didn’t have ONE methodology battle – they had THREE:
Methodenstreit | Era | Core Dispute
- First (Schmoller vs. Menger) | 1880s | Historical/inductive vs. theoretical/deductive method
- Second (Rieger vs. Schmalenbach) | 1927 | Economic efficiency vs. owner profitability
- Third (Gutenberg debate) | 1951 | Mathematical vs. empirical approaches
The Germans spent 60 years debating the PHILOSOPHY of what cost even means.
And they never fully resolved it. They just… kept debating. Because that’s what intellectually honest people do when confronted with genuinely hard epistemological questions.
The “Schlüsselproblematik” (The Allocation Key Problem)
And within all these theories lurked an even nastier problem: *Gemeinkosten* (overhead costs).
You have a factory roof. The IT department. Management salaries. Heating. How do you distribute these shared costs to individual products?
You pick a “key” (Schlüssel):
- Machine hours?
- Labor hours?
- Floor space?
- Revenue percentage?
- Headcount?
The choice is ARBITRARY. Different keys produce different “costs” produce different ROI produce different strategic decisions:
Same Product, Same Reality, Different Accounting:
Allocation by machine hours: Product A costs €47, ROI = 12%
Allocation by labor hours: Product A costs €63, ROI = -3%
Allocation by revenue share: Product A costs €51, ROI = 8%
Which is “true”? None. All. The question is meaningless.
Riebel’s Solution (That Nobody Used)
Paul Riebel proposed a radical solution in 1964: Relative Einzelkostenrechnung (Relative Direct Cost Accounting).
His “Identity Principle” (Identitätsprinzip) said: Only assign costs where you can prove direct causation. No arbitrary allocation. If you can’t trace it definitively, don’t allocate it.
The result was more honest. But also more uncomfortable – because suddenly you see how much of your “product cost” was actually fiction created by allocation keys.
Why didn’t Riebel win?
Because his system revealed that most “profitable” products were actually cross-subsidized by allocation artifacts. And nobody wanted to know that.
Also: it was hard to code.
Layer -1: The Philosophical Divide (German “Truth” vs. American “Usefulness”)
While Germans were debating epistemology, something very different was happening across the Atlantic.
The American Approach: Pragmatism Wins
In 1929, the stock market crashed. Millions lost everything. Congress investigated and found: accounting was a mess. Companies reported whatever they wanted.
Response:
- 1933-34: Securities Acts passed, SEC created
- 1936: “GAAP” terminology first used
- 1938: SEC delegates standard-setting to practitioners
- 1973: FASB takes over
The American question was never “What is TRUE?” but “What is USEFUL for investors?
The Continental Divide
Dimension | German Tradition | American Tradition
Core Question | “Was ist wahr?” (What is true?) | “What helps investors decide?”
Approach | Accounting as SCIENCE | Accounting as TOOL
Method | Theory-first, derive practice | Solve problems as they arise
Primary User | Creditors, tax authorities | Investors
Guiding Principle | Vorsichtsprinzip (Prudence) | Fair presentation
Asset Valuation | Conservative (lower of cost) | Fair value options
Tax Linkage | Strong (Maßgeblichkeit) | Separate
The Integration Gap
The German tradition produced *Betriebswirtschaftslehre* – an integrated theory of business economics that embedded accounting within a comprehensive framework.
The Americans never built this. Accounting stayed separate from management theory. As one accounting historian noted:
“There was no substantive emergence of a body of economic theory relating specifically to business in the English language area.”
Two completely different intellectual traditions. Two completely different answers to “What should accounting DO?”
And now they would collide.
Layer 0: SAP and the Murder of Theoretical Sophistication (1972-1984)
Five Engineers Leave IBM
In 1972, five former IBM employees in Mannheim, Germany founded SAP:
- Dietmar Hopp
- Hasso Plattner
- Klaus Tschira
- Claus Wellenreuther
- Hans-Werner Hector
They weren’t accountants. They weren’t theorists. They were software developers who wanted to build “real-time data processing” systems.
Their question wasn’t “Which Bilanztheorie is correct?” Their question was “What can we code?”
The ARIS Connection
Meanwhile, August-Wilhelm Scheer at Saarland University was developing ARIS (Architecture of Integrated Information Systems) – a framework for modeling business processes.
Scheer was a *Wirtschaftsinformatiker* (business information scientist), not a *Kaufmann* (business economist). His question wasn’t “What is theoretically correct?” but “How do we model processes for ERP implementation?”
And the simplest cost accounting model – Vollkostenrechnung (full cost accounting) with allocation keys – was by far the easiest to implement in software.
The Fateful Decision
When SAP built its CO (Controlling) module, it needed a cost accounting model. The choices:
- Deckungsbeitragsrechnung (Contribution Margin Accounting): Requires distinguishing fixed/variable costs. Judgment calls needed.
- Riebel’s Relative Einzelkostenrechnung: Requires proving causal relationships. Impossible to fully automate.
- Vollkostenrechnung with allocation keys: Pick a key, allocate mechanically. Deterministic. Programmable.
They picked the simplest option.
Not because it was theoretically superior. Because it was easiest to code.
The Sediment Formation
1960s: Multiple sophisticated cost accounting theories exist
Riebel’s Identity Principle offers philosophical rigor
↓
1972: Five IBM engineers want “real-time data processing”
Nobody asks “Which theory is correct?”
↓
1984: Scheer’s ARIS provides process models
Models assume SAP’s technical capabilities
↓
1990s+: 40,000+ enterprises adopt SAP
“Cost accounting” now MEANS “what SAP CO does”
↓
2000s+: Universities teach “SAP cost accounting”
Riebel becomes footnote in advanced courses
↓
Today: Nobody questions the CO engine
“That’s just how cost accounting works”
The philosophy that QUESTIONED whether objective costs exist is now buried under software that ASSUMES they do.
Layer 1: IFRS and the Anglo-American Hegemony (1973-2005)
The Origin (Not What You Think)
In 1973, the International Accounting Standards Committee (IASC) was established. The official story is “international harmonization.” The real story is more interesting.
A key impetus for establishing the IASC was British fear that the European Community would impose German-style codified accounting rules. The British wanted an “international” body to push Anglo-Saxon principles BEFORE the Germans could consolidate Continental influence.
The Voting Math
The IASC/IASB structure tells the story:
- Anglo-American bloc: US (5 votes) + UK (2) + Canada (1) + Australia (1) = 9 votes
- Required majority: 8 votes
- Everyone else: 5 votes
Nine votes. One more than needed. The Anglo-American bloc could pass anything unilaterally.
The “Harmonization” Results
Principle | Continental (German) | Anglo-Saxon | IFRS “Compromise”
- Primary user | Creditors | Investors | Investors ✓
- Philosophy | Prudence (Vorsicht) | Fair presentation | Fair presentation ✓
- Approach | Codified rules | Principles + judgment | Principles ✓
- Tax linkage | Strong | Separate | Separate ✓
- Asset valuation | Conservative | Fair value options | Fair value ✓
Notice the pattern? The “compromise” is identical to the Anglo-Saxon position.
The EU Mandate (2005)
In 2005, the EU mandated that all listed companies use IFRS for consolidated statements. This meant German companies – with 60 years of Bilanztheorie tradition – had to adopt Anglo-American principles labeled as “international.”
The philosophy that asked “Can we MEASURE true cost?” was replaced by the philosophy that said “Fair value equals market price or your model estimate.”
The Current German Mess
Post-IFRS, German companies now maintain:
- IFRS books (for investors)
- HGB books (for tax/legal compliance)
- Internal management accounting (for actual decisions)
Three parallel realities. None of them “true.”
Layer 2: Enron and Fair Value as Weapon (1992-2001)
The Philosophical Tool Becomes a Weapon
Remember the debate between German prudence (“Only book profits when REALIZED”) and Anglo-American fair value (“Book profits when ESTIMABLE”)?
In 1992, Jeffrey Skilling convinced the SEC to let Enron use “mark-to-market” accounting for energy contracts.
The SEC approved it. Because it was consistent with Anglo-American “fair value” philosophy.
The Mechanism
Traditional (German-style) accounting:
- Book revenue when EARNED
- Book costs when INCURRED
- Prudence: If in doubt, DON’T book profit yet
Mark-to-market (Anglo-American “fair value”):
- Sign a 20-year gas contract → Book ALL expected profits TODAY
- Estimate future cash flows using YOUR model
- Discount to present value using YOUR discount rate
- Record as current earnings
Mark-to-Market vs. Mark-to-Model
The critical vulnerability:
- When there IS a liquid market: “Mark to market” (observable prices)
- When there is NO liquid market: “Mark to model” (YOUR estimates)
Enron CFO Andrew Fastow exploited this ruthlessly. When asked where the numbers came from, Skilling told analysts they were “black box” numbers.
“Black box numbers.” That’s what happens when you let companies MODEL their own asset values.
The Results
The abuse of mark-to-market accounting, combined with Special Purpose Entities to hide losses, resulted in Enron’s reported revenues being 95% overstated.
$60 billion in assets. The seventh-largest company in America. Built on accounting philosophy that assumed companies would honestly estimate their own future profits.
What German Prudence Would Have Said
Schmalenbach, Riebel, and the entire German tradition would have responded:
“You cannot book profit from a 20-year contract on day one. The contract is not yet PERFORMED. The cash has not yet FLOWED. You have an EXPECTATION, not a REALIZATION.”
But the Anglo-American philosophy said: “If we can MODEL the expected value, we can book it.”
The “Fix” That Fixed Nothing
After Enron, America got Sarbanes-Oxley (2002):
- More auditor independence
- More internal controls
- More disclosure requirements
What they DID NOT fix: The fundamental assumption that “fair value” based on models produces OBJECTIVE numbers suitable for decision-making.
Layer 3: 2008 – Same Pattern, Different Instruments
Seven years later, the same pattern repeated.
Mortgage-backed securities. CDOs. “Mark to model.”
When there’s no liquid market for your exotic financial instrument, you use YOUR MODEL. And when your model says everything is AAA-rated and worth par, you book billions in profits.
Until reality arrives.
The 2008 financial crisis was, at its core, the same philosophical failure: the belief that MODEL-BASED VALUATIONS equal OBJECTIVE REALITY.
More regulations followed. More controls. More disclosure.
The philosophy remained untouched.
Layer 4: Today’s AI ROI – The Ultimate Epistemological Bankruptcy
And now we arrive at the present moment, where executives ask “What’s the ROI of our AI transformation?”
Let’s trace what that question actually involves:
The Archaeology of One AI ROI Number
You’re asking for: A single number that represents the return on your AI investment.
That number is calculated using: Cost accounting data from your ERP system.
That ERP system (probably SAP) uses: Allocation keys chosen decades ago by someone who may have left the company.
Those allocation methods reflect: A simplified model that five IBM engineers chose in 1972 because it was easiest to code.
That simplification ignored: 60 years of German debate about whether objective cost measurement is even POSSIBLE.
The “fair value” elements are based on: Anglo-American philosophy that “won” through voting majority, not epistemological superiority.
That philosophy enabled: The largest corporate fraud in American history (Enron).
The valuations of AI benefits use: “Mark to model” – YOUR estimates of future productivity gains.
And nobody in the boardroom knows ANY of this.
The ROI Calculation Stack
What they think they’re calculating: Objective financial mathematics
What they’re actually calculating:
- Layer 4: “AI benefits” modeled by consultants with skin in the game
- Layer 3: Using “fair value” assumptions that enabled Enron
- Layer 2: Under IFRS rules designed by Anglo-American majority
- Layer 1: With costs from SAP’s simplified CO module
- Layer 0: Based on allocation keys chosen for coding convenience
- Layer -1: Ignoring German philosophy asking if this is even possible
- Layer -2: On top of unresolved Bilanztheorien debates
- Layer -3: Using Brown’s diagnostic tool as a single-number oracle
The Specific Absurdities of AI ROI
1. The Productivity Gain Problem
“AI will improve productivity by 40%.”
Measured how? Against what baseline? In what timeframe? With what confounders controlled? Using whose productivity definition?
Most AI ROI calculations use consultant estimates of productivity gains. These consultants are paid to implement AI. Their models assume success.
2. The Attribution Problem
Your company implements AI. Revenue goes up.
Was it the AI? The new sales director? The competitor’s bankruptcy? The economy? The six other initiatives launched simultaneously?
Attribution in complex systems is essentially impossible. But ROI requires it.
3. The Timeframe Problem
ROI assumes you can specify a meaningful timeframe.
AI capabilities compound. Network effects emerge. Competitive dynamics shift. The “return” in year 3 depends on decisions that haven’t been made yet.
4. The Opportunity Cost Invisibility
ROI compares “investment vs. return.”
It doesn’t show: What else could you have done with that money/time/attention? The paths not taken are invisible to the calculation.
5. The Capability Blindness
Some investments create CAPABILITIES, not immediate returns.
“What’s the ROI of learning to walk?” is a meaningless question. The value is in what walking ENABLES, not in walking itself.
Much of AI investment is capability-building. ROI frameworks can’t capture this.
The Complete Sediment Stack
Let me lay it all out:
SURFACE LEVEL: “What’s the ROI of our AI transformation?”
(Spoken as if asking a mathematical question)
LAYER 4: AI “BENEFITS” CALCULATION (2024)
- Modeled by consultants with implementation incentives
- Based on assumptions about productivity no one can measure
- Using the same “mark to model” approach that built Enron
LAYER 3: FAIR VALUE PHILOSOPHY (Post-2008)
- Same assumptions that enabled 2008 financial crisis
- “Model-based valuation = objective reality”
– No fundamental philosophical repair
LAYER 2: ENRON WATERSHED (2001)
- Proved fair value philosophy could be weaponized
- “Fixed” with controls, not philosophy change
- Mark-to-model vulnerability remains
LAYER 1: IFRS ADOPTION (2005)
- Anglo-American philosophy labeled “international”
- German prudence tradition abandoned
- “Fair value” becomes global standard
LAYER 0: SAP CO MODULE (1972-1984)
- Engineers chose simplest implementable model
- Riebel’s sophistication rejected for coding convenience
- Arbitrary allocation keys hardcoded globally
LAYER -1: PHILOSOPHICAL DIVIDE (1900-1970)
- Germans: “Is objective cost measurement POSSIBLE?”
- Americans: “What’s USEFUL for investors?”
- Unresolved – just outvoted
LAYER -2: BILANZTHEORIEN WARS (1900-1960)
- Multiple competing theories
- Schlüsselproblematik never solved
- Riebel’s Identity Principle ignored
LAYER -3: DUPONT FORMULA (1914)
- Brown created DIAGNOSTIC decomposition
- Designed to ask “WHY is return X?”
- Now used as single-number oracle
THE FOUNDATION: THE ASSUMPTION NOBODY EXAMINES
- That costs are objective facts
- That they can be measured precisely
- That comparing them produces truth
- That a single number can capture complex reality
THIS ASSUMPTION IS CONTESTED AT EVERY LAYER. BUT NOBODY IN YOUR BOARDROOM KNOWS THAT.
What Can Be Done?
I’m not going to tell you to abandon ROI calculations. The world runs on them. Your budget approval process requires them. Your board expects them.
But I will tell you what HONEST practitioners do:
1. Know Your Sediment
Understand that your cost data is not objective fact. It’s the output of:
- Allocation keys someone chose years ago
- Software that encoded one simplified theory
- Philosophy that “won” by voting, not proof
When someone quotes a cost number to six decimal places, they’re performing precision theater. The first decimal place is probably fiction.
2. Decompose, Don’t Reduce
Use ROI the way Brown intended – as a diagnostic decomposition, not a single verdict.
“Our ROI is 15%” tells you nothing.
“Our margin is strong but asset turnover is killing us” tells you where to look.
3. Scenario, Don’t Point-Estimate
Never present a single ROI number. Present ranges:
- Pessimistic: What if nothing works as planned?
- Expected: What reasonable assumptions suggest
- Optimistic: What perfect execution might achieve
If your scenarios don’t span a wide range, you’re lying to yourself about your certainty.
4. Name Your Assumptions
Every ROI calculation contains assumptions. Most are hidden. Make them visible:
- “This assumes productivity gains of 40% based on [source]”
- “This uses allocation keys from 2018 based on [rationale]”
- “This timeframe was chosen because [reason]”
If you can’t name your assumptions, you don’t understand your number.
5. Track Actuals Against Predictions
The greatest lie in corporate finance is that nobody goes back to check.
Build the discipline: After 2 years, did the AI investment deliver what the ROI calculation predicted? If not, why not? What does that teach you about your modeling?
Organizations that do this discover their models are terrible. Then they improve.
6. Recognize the Complex Domain
Most “transformation” work is in the Complex domain (Cynefin framework). In complex systems:
- Outcomes emerge – they can’t be predicted
- Causation is entangled – attribution is impossible
- Learning happens through probing – not planning
Asking “What’s the ROI?” of a complex initiative is like asking “What’s the ROI of having children?” The question misunderstands the nature of the endeavor.
Epilogue: The Honest Answer
The next time someone asks you “What’s the ROI?”, you have several options:
Option A (Career-Limiting): “That question assumes costs are objective facts, that future benefits can be modeled, that attribution in complex systems is possible, and that a single number can capture multidimensional value. Each of those assumptions is contested by 111 years of accounting philosophy. Which assumption would you like me to pretend is true?”
Option B (Diplomatic): “I can give you a range based on explicit assumptions. But more usefully, let me show you the key uncertainties and what would have to be true for this to succeed.”
Option C (Honest): “I don’t know. Neither do you. Neither does anyone. What I can tell you is what we’ll learn by trying, and what signals will tell us we’re on track or off.”
The ROI incantation works because it provides false certainty in a world that offers none. It converts messy reality into clean numbers. It transforms irreducible complexity into boardroom-ready slides.
And it’s built on 111 years of unresolved philosophical debate, simplified software decisions, contested political victories, and at least two moments where the foundational assumptions enabled catastrophic fraud.
When someone says “ROI” like they’re stating a fact, they’re actually standing on an archaeological dig site, speaking with confidence about ground that shifts all the way down.
Now you know what’s beneath your feet.
Sleep well.