Insights · Learn
Regret isn't a feeling. It's a structural condition.
When a position underperforms and the team reopens the IC memo, 'was this a good decision' often degrades into 'who do we blame.' That degradation is structural, not personal.
Deal memo regret happens when a team reopens an old IC memo after a position underperforms and 'was this a good decision' turns into 'whose fault is this.' That shift is structural, not personal: the memo decays into a static document, the people who knew the details leave, evidence fragments across emails and drives, and hindsight bias rewrites what the team believed it knew. It is what happens when decision records aren't built to stay connected to their evidence.
Why this matters
Institutional firms outlast the individuals working in them. Partners retire, associates leave, leadership transitions. Each transition takes a slice of institutional decision memory with it: what the team actually knew at approval, what assumptions were tested, what conditions were committed to monitor, what alternatives were rejected and why.
When a position underperforms years later, the natural question is 'what did we know at the time?' The structural problem is that the people who knew are often gone, and the artifacts they left behind were built to communicate a decision in the moment, not to answer that question later. The question degrades into 'whose fault is this,' which is about personalities rather than process.
The five structural sources of deal memo regret
The memo decays into a static document
A 30-60 page memo at approval goes out of date within months; updates happen informally, not in the memo itself.
Team turnover walks out with institutional memory
The analyst, the partner and the operating lead each take working knowledge the artifacts did not capture.
Evidence fragments across emails, decks and drives
The model, the QoE and the diligence files end up scattered, and reassembling them years later is error-prone.
Hindsight bias rewrites what the team knew
Assumptions that were reasonable at the time read as obviously flawed once the outcome is known.
No anchored record at decision time
Without a fixed record of what was explicitly tested versus implicitly accepted, the team is left relying on memory.
A typical IC memo runs 30-60 pages of structured analysis at the moment of approval. Within six months the operating environment has shifted and new information has emerged, but the team rarely updates the memo itself; narrative updates happen instead in quarterly reports, board materials and informal commentary. Years later, the question 'what did the IC actually approve' can still be answered from the memo. The question 'was the decision still defensible six, twelve or eighteen months in' cannot, because the memo was not built to answer it.
Team turnover compounds the gap. The analyst who underwrote the deal moves to another firm, the partner who approved it retires, the operating lead who managed the position transitions to a different company. Each exit removes a slice of working knowledge that did not make it into any document: the alternatives the team debated and rejected, the conditions it committed to monitor informally, the context behind a number that looked unusual at the time. The replacement team inherits the artifacts, not the working knowledge behind them.
The surviving artifacts are rarely in one place either. The model sits in a shared drive folder, the QoE in an email attachment, the diligence files in a data room that may have expired, the management presentations on someone's laptop. Reassembling the full decision basis years later means tracking down each fragment, which is slow, error-prone, and sometimes impossible if even one fragment went missing along the way.
The 'we already underwrote this' pattern
A specific version of this is worth naming. A new partner reviews a struggling position and proposes intervention. A senior team member says 'we underwrote this, customer concentration is a risk we accepted at IC.' The real question is whether that risk was accepted with explicit downside testing and committed monitoring, or accepted implicitly because nobody pressed on it. The two are very different commitments, and without a fixed record, memory tends to expand favorably for whoever is making the claim. A risk that was explicitly stress-tested and a risk that simply went unchallenged end up sounding identical in hindsight, which is precisely the structural trap.
A risk that was not explicitly stressed gets remembered the same way as one that was stressed but at too little confidence. Merging the two in hindsight is a memory failure, not bad faith.
What keeps a decision record connected to its evidence
The structural fix is not better software for writing memos faster. It is keeping the record connected to the evidence it was built from, so reopening a position does not require reconstructing the decision from memory. That takes three things: a fixed record of what was known at approval, a way to see what has changed since, and a chain connecting approval through to any later intervention and outcome, none of which can depend on whoever happens to still be at the firm.
A memo in a folder
- Figures re-keyed from the spread by hand
- Risks written as prose only
- Conditions buried in the approval email
- Reassembled from scratch each time someone asks
A memo built from the record
- Figures trace to the document they were spread from, where a locator resolves
- Risk signals graded against the firm's own thresholds
- Open conditions kept on the same deal record after close
- Exported as DOCX or XLSX with a deterministic evidence fingerprint

What this means for LP review and audit
A well-connected decision record earns its value when someone outside the deal team asks a hard question. Three come up repeatedly, and each needs a different piece of the record to answer.
| Who asks | The question | What answers it |
|---|---|---|
| An LP | When did you first know things were going badly? | A dated log of what changed against the thesis the IC approved |
| An auditor | How did the board exercise oversight on this position? | A timeline connecting approval, monitoring and any intervention decisions |
| A future buyer | What was the original thesis, and how did it evolve? | The chain of decisions, with the evidence each one was based on |
None of those answers should require the original analyst to be in the room, and none should depend on a partner's memory of a meeting from three years earlier.
What this changes for IC governance
- Decision quality becomes evaluable separately from decision outcome; see decision quality vs decision outcome for the framework.
- Post-mortems lean on the spread and the memo's sourced evidence rather than on analyst memory.
- A new team member inherits a memo traced to its sources, not just a document to take on faith.
- Board and LP review can point to conditions that were recorded on the deal record, not reconstructed from an approval email.
- The firm's working knowledge stops living only in the heads of whoever happened to run the deal.
The argument mirrors the case Atul Gawande makes for checklists: judgment plus structural infrastructure consistently beats judgment alone under pressure. Firms that produce top-quartile outcomes across many cycles are usually not the ones with the most talented individual dealmakers; they are the ones whose institutional discipline survives the people who built it moving on.
The cure for regret is structural, not personal: keep the memo connected to the evidence it was built from, so a future reviewer does not have to trust memory.
Sources cited
- Daniel Kahneman, Thinking, Fast and Slow (Farrar, Straus and Giroux, 2011). Hindsight bias and System 1/System 2 cognition.
- Daniel Kahneman, Olivier Sibony and Cass R. Sunstein, Noise: A Flaw in Human Judgment (Little, Brown Spark, 2021). Institutional decision discipline.
- Philip E. Tetlock and Dan Gardner, Superforecasting (Crown, 2015). Calibration and the cost of failing to anchor predictions.
- Annie Duke, Thinking in Bets (Portfolio, 2018). The 'resulting' framework and the cost of conflating decision quality with outcome.
- Michael J. Mauboussin, The Success Equation (Harvard Business Review Press, 2012). The luck-skill continuum.
- Atul Gawande, The Checklist Manifesto (Metropolitan Books, 2009). Structural discipline as infrastructure.
Questions
Why does IC memo regret feel personal when it's actually structural?
Because the degradation from 'was this a good decision' to 'whose fault is this' happens quietly, through memo decay, team turnover and hindsight bias, rather than through any one person's error. It reads as personal because the result lands on whoever is in the room, not on the structural gap that produced it.
What is the 'we already underwrote this' pattern?
It is when a team cites a risk as already accepted at IC, without a clear record of whether that risk was explicitly stress-tested or simply went unchallenged. Both get remembered the same way without a fixed record of which one actually happened.
How does hindsight bias affect deal post-mortems?
It makes a reasonable assumption look obviously flawed once the outcome is known, and a noted-but-discounted risk look obviously dispositive. The discipline is to judge the decision against the evidence available at the time before layering in the outcome.
Does team turnover really remove decision-making knowledge?
Yes. The analyst who underwrote the deal, the partner who approved it, and the operating lead who managed it each carry working knowledge the memo and model were not built to capture, so a replacement team inherits the documents but not the context behind them.
Can a decision record actually prevent regret, or just explain it afterward?
It can only prevent the memory failure, not the bad outcome itself. A record that stays connected to its evidence lets a later reviewer check what the team actually knew at approval, instead of relying on recollection that tends to shift once the result is known.