How Products Fail — a TaxonomyStart here

Across every case study on this site, the failures fall into six patterns. Naming them is what lets you recognise the seventh before it has a name.

Why a taxonomy rather than a list

• Every event in the case studies here was described at the time as unprecedented. Almost none of them were.
• The instruments change, the jurisdictions change, the decade changes. The mechanism repeats, and there are roughly six of them.
• The practical use: when you meet a product you have not seen before, you cannot look up its history. You can ask which of the six it is exposed to, and that question is answerable from the term sheet.

1. Leverage plus a mark-to-market obligation

• The mechanism: the position is right, the financing is short-term, and the margin call arrives before the thesis pays off. Being early is indistinguishable from being wrong when someone else controls the timing.
• Where it shows up: LTCM, Metallgesellschaft, Archegos, LDI in 2022.
• The tell: a hedge whose cash flows do not match the thing it hedges in timing, only in size. Metallgesellschaft's hedge was economically sound and cash-flow fatal.
• What to measure: the distance to the call, not the distance to being wrong — the margin-call simulator.

2. Correlations converging under stress

• The mechanism: the diversification was measured in calm conditions. Under stress, everything that is being sold by the same forced sellers moves together, and the portfolio turns out to have been one position.
• Where it shows up: 2008, 1998, and inside every "diversified" book that was diversified by name rather than by risk factor.
• The tell: a correlation assumption estimated on a sample that contains no crisis.
• What to measure: the same allocation with all pairwise correlations forced to 0.8, in the three-asset risk tool. If that result is unacceptable, the portfolio is a bet on calm.

3. Liquidity mismatch between the wrapper and the holdings

• The mechanism: daily dealing on assets that take weeks to sell. It works until redemptions arrive, at which point the first sellers are paid with the liquid assets and the remaining holders own the residue.
• Where it shows up: property funds gating in multiple jurisdictions and decades, credit funds in March 2020, and the structural design of any open-ended vehicle over illiquid assets.
• The tell: the redemption terms and the asset liquidity described on different pages of the same document, never side by side.
• Why it is not solved: swing pricing and notice periods reduce the transfer between holders; they do not remove the mismatch, because the mismatch is the product.

4. Counterparty and issuer failure

• The mechanism: the payoff was right and the entity owing it was not there. The market view becomes irrelevant.
• Where it shows up: FTX, structured notes from failed issuers, uncleared derivatives in 2008, and any position where "safe" described the payoff rather than the promise.
• The tell: an instrument described by what it tracks, with the entity that owes it named only in the small print. See investor protection.
• The structural fix — segregation, clearing, collateral — works, and is precisely what distinguishes a fund from a certificate with identical exposure.

5. Model failure: the map was not the territory

• The mechanism: a valuation, a risk number or an index rule assumed something that stopped being true. The number was not wrong until the assumption was.
• Where it shows up: negative oil prices in 2020 (the model could not represent them), Volmageddon (the rebalancing rule became its own driver), correlation models in 2007.
• The tell: a bound that is assumed rather than contractual. "Prices cannot go below zero" was an assumption; the contract said no such thing.
• What to do: read the mechanical rule, not the description of the rule. A leveraged product's daily reset is a rule; "tracks twice the index" is a description — see leveraged ETPs.

6. Documentation risk

• The mechanism: the payoff worked exactly as written, and what was written differed from what everybody assumed.
• Where it shows up: the 2023 AT1 write-down, where a clause in the terms inverted the expected creditor hierarchy; barrier definitions where American and European mean entirely different products; "capital protected" applied to a note rather than a deposit.
• The tell: a widely shared assumption about how something "obviously" works, which nobody has traced to a sentence in the terms.
• What to do: the term-sheet playbook exists for this failure mode specifically.

The pattern across all six

• None of the six is a market-direction problem. In every case the loss came from structure, timing or wording rather than from the view being wrong.
• All six are visible before the event, in documents that are public at the time of purchase. That is the uncomfortable conclusion and the reason this site is built around reading documents.