What past failures teach about building a decision model

What do past failures teach about building a decision model?

Past failures teach one practical rule: build the model around signals that were visible before the collapse. In each public case, the warning sat in documents, filings, or the absence of evidence. A good model reads those traces systematically and turns them into a score, so the alarm does not depend on one analyst noticing the right page on the right day. Each lesson below comes from a public case, with its date and its source.

The table below maps each case to the signal family it teaches and to the dimensions that carry it. The EQOS Asia method scores 9 structural dimensions and returns a φ(S) value between 0 and 1, with a STOP zone below 0.618 (EQOS Asia method page, as of October 2026).

Why do public failures make good test material for a decision model?

Public cases offer three assets at once. The facts existed on a known date, the outcome is known today, and any reader can check the sources. That combination lets a model be tested on history, with each input traceable. The EQOS Asia case studies state that their back-tests use data available before the event (EQOS Asia case-study pages, as of October 2026).

The EQOS Asia pages label these studies as retrospective educational exercises, and they do not constitute investment advice. That framing keeps the lesson useful: the reader learns which signals to look for, and the model shows how to weigh them. For a decision of your own, the method page explains how the Audit, Modelling, and Prediction pillars apply.

What does the 2008 crisis teach about decision models?

The EQOS Asia back-test applies the Codex to public data from 2006 and 2007 and returns φ(S) = 0.234, with 6 of 9 dimensions critical (EQOS Asia 2008 case page, as of October 2026). The lesson for a model is that leverage, opaque structures, and conflicts of interest are measurable signals that can be scored before a shock.

The dedicated guide on what the 2008 crisis teaches follows these signals one by one. A key point is that energy can mislead. Record profits sat beside very low integrity, and the aggregation lets the weakest dimension pull the total down, so a model that respects the weakest link reads fragility where an average reads health.

Which warning signals does the Theranos case teach a model to score?

The Theranos back-test returns φ(S) = 0.085, presented on that page as the lowest EQOS Asia score computed, with 8 of 9 dimensions critical (EQOS Asia Theranos case page, as of October 2026). Its main signal is an absence: the page records 0 peer-reviewed publications over 10 years of research, and absence of evidence becomes an input the model scores.

The Theranos warning signals also include a board whose expertise sat far from the mission and a technology kept behind closed doors. Each maps to a dimension: alignment between competence and purpose, and transparency about the product. The guide shows how a model sequences these checks, from the cheapest to the costliest.

Which valuation signals does the WeWork case teach a model to score?

The WeWork back-test returns φ(S) = 0.189 on public 2017 data, with 7 of 9 dimensions critical (EQOS Asia WeWork case page, as of October 2026). The page compares revenue with lease commitments: US$1.8 billion of revenue against US$47.2 billion of obligations gives a ratio of 0.04, far under the viability threshold of 1.5 stated there.

Two more signals sit in the same page: super-voting shares at 20 to 1, and an adjusted profit metric that diverged from the reported loss. The WeWork valuation signals guide explains how a model compares the valuation multiple of a company with the multiple of its real business model.

How do you read a back-test, and what do startup failures add?

A back-test gives evidence about a method, so a careful reading comes first. The checklist covers input dates, replay, dimension drivers, limits, and thresholds. The guide on how to read a back-test result gives each step, with the questions to ask before trusting a score. A back-test checks a method, and a decision still needs the inputs of your own case.

Startups add a second lens, because their failures rarely come from one cause. The sector page scores startups at φ(S) = 0.333 and recommends measuring a startup before investing or scaling (EQOS Asia startups case page, as of October 2026). The guide on startup failure signals lists the checks a decision engine runs first.

Can a model spot a bubble, and where do you start?

A model can spot a bubble when it compares enthusiasm with transparency. On the 2008 data, the EQOS Asia page puts resonance at 0.65 against transparency at 0.15, a ratio of 4.3 to 1 that exceeds its 3 to 1 anomaly line (EQOS Asia 2008 case page, as of October 2026). A flag raised in time lets a decision maker pause, request evidence, and set conditions.

To test your own decision, start with one question and nine ratings. The free simulator estimates φ(S) in about 10 minutes, and a 30-minute video call applies the EQOS Asia equation to a real case, scheduled in Singapore time (EQOS Asia simulator and contact pages, as of October 2026).

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile