Introduction
How to read this book
How to read this book
1. One argument, thirteen chapters#
Every edge that can be bought is bought by everyone. What remains is how a firm decides, and this book is about making that decision the best it can be: the case for it, the framework behind it, the system that runs it, the clients it serves, the practice that delivers it and the language it uses. Each chapter is a complete paper that can be read on its own, so a reader who starts in the middle loses nothing; read in order, the chapters build one argument.
An edge that can be bought is bought by everyone.
Data, models and screens are commodities, and a signal edge decays once it is found. What remains is how a firm decides.
The decision is the unit of alpha.
A decision is the commitment to an action, followed through to the next action. It is judged by its quality, not by one outcome: calibration, policy adherence and conviction against outcome separate skill from luck.
Four inefficiencies, beyond human limits.
Every firm suffers the same inefficiencies in governance, information, analysis and behaviour. They come from uncertainty, complexity and speed, not from a weak team, and they cost measurable monetary value.
A good decision is judged against the worlds that did not happen.
History is one path among many. Causal and counterfactual reasoning, not a bigger dataset, a retrained network or a language model, is what a high-stakes decision needs; quant, quantamental, deep learning and language-model strategies each stop short of it.
Decision optimization = Strategy + Intelligence + Platform.
The strategy is the firm's own. The intelligence runs Perceive, Diagnose, Predict, Recommend in every analysis. The platform runs it, governs it and learns from it.
Four levels, four policy types, eleven decision archetypes.
A solution is the decision lifecycle of one domain, a decision policy one decision formalized, an analysis one question answered, an analysis archetype one reusable method. Solutions are assembled from decision assets, so the same decision runs the same way every time.
DeOS is infrastructure, not a tool.
A dedicated instance with five capabilities, Governance, Context, Analysis, Recommendation and Learning, that the firm owns, secures and builds on, with technology built for decisions rather than for text.
Value is underwritten, proven and scaled.
Every decision archetype and segment carries a value equation in monetary value. A CIO evaluates at three levels, never on one outcome, and delivery runs in six phases with four gates, from Discovery to Scaling.
Reading paths
Investor or board member
CEO or CFO of a company
CIO or head of strategy
Quant or research lead
Analyst, data or technology lead
Consultant or decision architect
Chapters 1, 2 and 10, then the segment page that fits in Chapter 9.
Chapters 1 and 2, the enterprise version of the questionnaire in Chapter 12, then Chapter 9.
Chapters 1 to 5, then 7, 9 and 10; Chapter 6 when the first solution is designed.
Chapters 3, 4 and 8 first, then 5 to 7; Chapter 13 as the reference.
Chapters 5 to 8; Chapter 13 as the reference.
Chapters 5, 6, 11 and 12 end to end; Chapter 9 for the roadmap and the catalogue; Chapter 13 always at hand.
Conventions#
Terms are used as the glossary (Chapter 13) defines them, and where the papers once differed the glossary states the convention this book follows. Numbers marked illustrative are examples, not client data. Sources are cited in Chapters 4 and 8, which carry the comparison and the technology claims; the other chapters cite within the text only where a claim rests on a named study.