Investment mistake taxonomy across four levels
Investment mistake taxonomy across four levels
One-line summary: Josh Steiner — co-author of From Mistakes to Meaning and decades-long Treasury/IB/PE/IC operator — partitions investing mistakes into four distinct categories, each with different error structures and learning loops: deal-level, investment-business-management, people-management, and IC-service.
What it is
The Capital Allocators interview (Ep. 502) is structured around Steiner walking through mistakes across four levels of the investing operation, then closing with frameworks for avoidance. The taxonomy implicitly argues that the kind of mistake determines whether the learning loop closes: deal-level mistakes are fast-feedback (you know within a few years), management mistakes are slow-feedback and laden with motivated reasoning, IC-service mistakes are very slow and structurally hard to attribute. The implication for practitioners is that the categories cannot be lumped — a mistake-tracking system that doesn't separate them will systematically over-learn from deal mistakes and under-learn from management/governance mistakes.
The four categories:
- Deal-level mistakes — wrong company, wrong price, wrong thesis-falsifying evidence ignored. Fast feedback (1-3 years to resolution typically).
- Investment-business-management mistakes — building the wrong organization, wrong incentives, wrong infrastructure to support repeatable judgment. Medium-slow feedback (3-7 years).
- People-management mistakes — wrong hires, wrong promotions, wrong delegation patterns. Slow feedback; bias-vulnerable.
- Investment-committee service mistakes — the IC seat is a particular kind of authority-and-responsibility position where the dynamics of group decision-making, time-budget, and information access differ from being a principal at a fund. Very slow feedback; often invisible because the IC member doesn't own the consequences directly.
Why it matters to psychology
Each category has different cognitive failure modes:
- Deal mistakes are vulnerable to confirmation bias and motivated reasoning on entry.
- Business-management mistakes are vulnerable to organizational identity preservation — once you've built the org around a thesis, you can't unwind without admitting the larger error.
- People-management mistakes are vulnerable to sunk-cost fallacy and attachment — the literature on persistent hiring errors maps cleanly here.
- IC-service mistakes are vulnerable to diffusion of responsibility and epistemic humility miscalibration (deferring too much OR not deferring enough to the principal).
The bigger psychology point: feedback latency drives learning quality. Mistakes with fast, undeniable feedback (deal mistakes) get learned; mistakes with slow, ambiguous, or attributable-elsewhere feedback (IC, management) do not — even when the actor is a deliberate, reflective learner like Steiner.
Evidence
- josh-steiner in 2026-05-18-podcast-capital-allocators-making-mistakes-josh-steiner-ep-502: Episode is structured around the four categories explicitly, with Steiner walking through one major mistake of each kind he has personally made over 30+ years across Treasury, Quadrangle Group (PE), Bloomberg (operator), Yale Investment Committee (allocator), and SSW Partners (PE founder).
- josh-steiner in same, on the Treasury mistake: kept a personal diary that got subpoenaed under the Whitewater investigation. "For anyone who's ever kept a journal or a diary or written a letter to a girlfriend, you're not trying to provide a verbatim account of what happened. You're trying to write impressionistically. And in my case... what I had written didn't completely comport with what I remembered because it was an attempt to write a verbatim account." The mistake at issue was a process-and-disclosure one — and the cognitive lesson was that impressionistic personal records will not survive an evidentiary frame later applied to them. This is a category Steiner doesn't fit cleanly into the four-level taxonomy: it's neither an investing mistake nor a leadership mistake, it's an evidentiary-self-presentation mistake. Worth noting as a fifth category the wiki may want to track separately.
Implications
- For mistake-tracking systems: separate the four categories before counting incidents; otherwise the deal-mistake feedback drowns out the management/IC signal.
- For IC composition decisions: IC-service mistakes' feedback latency means the selection of IC members has to substitute for the learning that other roles afford — once on the IC, you don't get many chances to update before tenure ends.
- For book-writing about mistakes (the explicit goal of Steiner's book From Mistakes to Meaning): the taxonomy implies you can't tell one mistake story; you have to walk the levels, because the lessons don't transfer across them.
Open questions
- Does the taxonomy survive operator-vs-allocator-vs-PM specialization, or do those roles have different mistake-levels altogether?
- Where do public-private interface mistakes (Steiner's Treasury → finance pivot is loaded with these) fit — are they a fifth category or a hybrid?