An investment narrative can explain why a position makes sense. It is much worse at telling a team when that view should change. Before capital and identity attach to a thesis, decompose it into assumptions, evidence tests, and explicit rules for updating conviction.
“Is the thesis still intact?”
It is one of the most common questions in portfolio reviews, and one of the least precise.
A company misses a quarterly target but maintains guidance. A private investment gains customers more slowly than expected but improves unit economics. A macro position moves against the portfolio while one of its underlying indicators strengthens. The team gathers, revisits the original rationale, and asks whether anything fundamental has changed.
Usually, the answer arrives as another narrative.
The analyst explains why the disappointing evidence may be temporary. The portfolio manager introduces a new development that supports the longer-term view. Someone observes that valuation has become more attractive. By the end of the discussion, the original thesis may sound different from the one that justified the investment, yet the conclusion remains the same: intact.
The problem is not that stories are inherently bad. Investing requires causal reasoning, and a coherent narrative is one of the best ways to communicate a complex view.
The problem is using the narrative as the monitoring system.
Stories are built for coherence. Monitoring systems need to preserve uncertainty, contradictions, and changes of mind. A thesis that exists primarily as prose can absorb new information almost indefinitely because the explanation can be rewritten faster than the investment can be falsified.
A better discipline is to treat the thesis as what it really is: a set of hypotheses about how the world works.
Before committing capital, teams should separate those hypotheses into their critical assumptions, assign an explicit level of confidence, identify evidence that would support or weaken each one, specify leading signposts and review thresholds, and record who is responsible for revisiting them.
The practical output is not another investment memo. It is a living thesis register.
Its purpose is not to mechanize portfolio management. It is to make changes in conviction visible before hindsight, price moves, and ownership of the position rewrite the reasons the investment was made.
Narratives explain. Hypotheses can be tested.
An investment narrative compresses a great deal of reasoning.
A software company will sustain growth because its product is becoming embedded in customer workflows. A roll-up will create value because the fragmented market permits continued acquisitions at attractive multiples. A bank will rerate because credit normalization will reveal structurally higher earnings power.
Each statement can be perfectly sensible. Each also hides several separate propositions.
For the software company, customer dependence must actually be increasing. Higher dependence must translate into retention or pricing power. Competitive products must not erase the advantage. The resulting economics must persist long enough to justify the valuation paid.
If those propositions remain bundled inside one story, evidence against one can be offset rhetorically by evidence for another. Retention weakens, but new-logo growth is strong. Pricing disappoints, but usage is rising. The original source of value can quietly migrate.
This is not merely a problem of poor discipline. Commitment itself can affect subsequent judgment.
In Barry Staw’s classic experiment on escalating commitment, 240 business students participated in a simulated business investment decision. Participants who were personally responsible for an earlier choice and then received negative results subsequently committed more resources to that course of action than other groups. The experiment was stylized and should not be treated as a direct model of professional portfolio managers, but it demonstrated a mechanism investment organizations should take seriously: adverse evidence does not necessarily weaken commitment when the decision maker owns the prior choice.
That is why “remain intellectually flexible” is weak advice. Flexibility is hardest precisely when it is needed.
A more robust response is structural: externalize the assumptions before ownership hardens.
MIT CISR makes a similar distinction in its work on organizational learning. Ross and Fonstad argue that large initiatives often embed a whole set of assumptions inside what appears to be one hypothesis. Breaking those assumptions apart makes it possible to conduct smaller tests that confirm or disconfirm specific beliefs rather than waiting for the entire initiative to succeed or fail.
The same principle applies to investing.
The question is not merely, “Is our thesis right?”
It is:
Which propositions have to be right for the investment to work, and what evidence would cause us to change our confidence in each one?
Decompose the thesis before commitment
The most important work on a living thesis happens before the trade, acquisition, or funding decision.
At that point, the team still has an unusual advantage: it can define what future evidence should mean before knowing what that evidence will be.
Start with the causal chain.
Suppose the proposition is that a company’s margins will converge toward peers. That is a forecast, not yet a thesis. The causal claim might be that current margins are depressed by investments whose costs will grow more slowly than revenue. That explanation in turn depends on assumptions about customer acquisition, retention, pricing, support costs, operating leverage, competitive responses, and management’s willingness to let incremental revenue reach the bottom line.
Those assumptions do not deserve equal attention.
Some may be observable and low consequence. Others may carry most of the expected return. A useful thesis register therefore forces the team to identify the small number of assumptions for which being wrong would materially change valuation, position size, or the decision to own the asset at all.
The exercise also separates the inside view from the outside view.
Kahneman and Lovallo describe the inside view as forecasting from the details and scenarios of the case itself, while the outside view begins with the outcomes of a relevant reference class. Their argument originated in project forecasting rather than securities analysis, but the mechanism transfers naturally: an especially persuasive company-specific narrative should not make the investor forget what has historically happened to comparable businesses, transactions, technologies, or turnarounds.
A useful pre-investment record might therefore distinguish between:
- What must be true: the causal assumptions that drive the investment.
- What is currently observed: facts available today rather than forecasts or interpretations.
- What the base rate says: relevant outcomes outside the specific case.
- What would discriminate: future evidence that should look different if the assumption is wrong.
- What the investment implies: how changes in the assumption should affect expected return, valuation, sizing, or ownership.
This is more demanding than attaching a “key risks” section to the end of an investment memo.
A risk factor can exist without threatening the causal thesis. A falsifiable assumption must connect directly to the logic that generates the return.
That distinction also reinforces a basic professional obligation. CFA Institute’s current Standard V(A) requires members and candidates to exercise diligence, independence, and thoroughness and to maintain a reasonable and adequate basis supported by appropriate research and investigation for investment recommendations and actions. It does not prescribe a thesis register, but the standard emphasizes process and evidentiary basis rather than confidence alone.
The practical implication is simple: conviction should be decomposable.
If no one can identify the assumptions whose failure would reduce conviction, “high conviction” is describing an attitude, not a research conclusion.
Decide what new evidence should mean before it arrives
The most difficult part of updating is not collecting new information. It is deciding how much weight the information deserves.
Investors face two symmetrical errors.
One is rigidity: the team explains away every adverse observation because a quarter is noisy, management needs more time, the market misunderstands the opportunity, or the original thesis is “long term.”
The other is hyperreactivity: every data release, channel check, price move, or management comment causes conviction to lurch.
Research on real-world forecasting illustrates the balancing problem. Atanasov, Witkowski, Ungar, Mellers, and Tetlock analyzed more than 400,000 probability forecasts covering almost 500 geopolitical questions in a four-year forecasting tournament. More accurate forecasters tended to update frequently but in relatively small increments, while weaker forecasters were more prone either to confirm their prior views or to make infrequent, large revisions. The domain is forecasting rather than investing, and association should not be confused with a universal trading rule. But the finding is useful because it rejects both caricatures of good updating: neither stubborn conviction nor constant reversal performed best.
Investment teams need the equivalent of an update function.
For each critical assumption, specify a leading signpost: an observable that provides information about the underlying mechanism before the final financial outcome is known.
Then distinguish a signpost from a threshold.
A signpost might be customer retention. The hypothesis concerns whether product integration is deepening. The team should decide in advance what pattern in retention would materially change its confidence: one weak monthly observation may be noise; deterioration across multiple cohorts, accompanied by a specific change in customer behavior, may be evidence against the mechanism.
The threshold is therefore not simply “retention below 90%.”
It is an answer to a harder question:
What amount and quality of evidence would be surprising enough under our current hypothesis that we should reconsider it?
McKinsey makes a related recommendation in its work on strategy under uncertainty: organizations should track the assumptions underlying a plan over time and establish explicit trigger points for revisiting decisions as new information becomes available. The authors note that assumptions often fade from memory even while variance against budgets remains closely monitored.
Investment teams should go one step further and distinguish an evidence threshold from a decision threshold.
The first says: this evidence is strong enough that our belief should change.
The second says: conviction has changed enough that the portfolio decision should change.
Those are not the same thing.
A thesis can weaken while a security simultaneously becomes cheaper. A negative operating development may reduce estimated intrinsic value but leave expected return attractive after a much larger price decline. A venture thesis can weaken without creating an opportunity to exit. A private-equity team may learn that one underwriting assumption is wrong while determining that a different value-creation path is sufficiently attractive to justify continued ownership.
The register should make those changes visible rather than forcing every piece of information into a binary “hold or sell” trigger.
A threshold should trigger judgment, not replace it
This distinction addresses the strongest objection to a formal thesis register.
Many important drivers are not measurable with precision. Data arrive with error. Leading indicators change meaning across regimes. Management teams alter behavior. Markets incorporate information that the investor cannot directly observe. Strict rules can make an organization act confidently on weak signals simply because a number crossed a line.
That would substitute false precision for narrative flexibility.
The answer is to design thresholds as review obligations, not automatic commands.
Crossing a threshold should require the team to reopen the relevant assumption, examine whether the signal is reliable, search for corroborating and disconfirming evidence, and explicitly record the resulting change, or non-change, in conviction.
Some evidence will justify a large revision. Most will not.
The register should also distinguish between what is unknown and what is unknowable at useful precision. Forcing a numeric probability onto every qualitative judgment can produce the appearance of rigor without additional information.
Probabilities are most helpful where the team can define the event clearly and expects to revisit the estimate. Elsewhere, calibrated confidence bands may be more honest. The important discipline is preserving the prior judgment so that the direction and magnitude of the update remain observable.
A register is therefore not an algorithm for investing.
It is a constraint against unconsciously changing the rules after observing the result.
Do not let a broken thesis become a new thesis without saying so
The most revealing moment in a portfolio review often occurs when an original assumption fails but the position still appears attractive.
Imagine that an investment was initially underwritten on accelerating market growth. Growth disappoints. But margins improve substantially faster than expected, and the valuation falls.
It may be perfectly rational to continue holding, or even to add.
But the team should not describe the original thesis as “intact.”
It has discovered a different investment.
This distinction matters because otherwise a portfolio can accumulate what might be called thesis drift: the reasons for owning an asset continuously change while the organization records an uninterrupted line of conviction.
The danger is not changing your mind. Changing your mind is the point.
The danger is changing the rationale without recognizing that a fresh underwriting decision has occurred.
A living register should therefore contain a status beyond “confirmed” and “disconfirmed”: re-underwrite.
That status is appropriate when a critical link in the causal chain has failed but new information creates a materially different source of expected return.
The team then asks the question it would ask if it did not already own the position:
Given what we now know, would we initiate this investment today, for these new reasons, at this price and size?
That question cannot completely neutralize ownership effects. But it makes the transition from old thesis to new thesis explicit.
This is where an outside view becomes useful again. When the causal explanation changes, the relevant reference class may change with it. A growth investment turning into a margin-recovery investment should not automatically inherit the base rates, horizon, or risk budget of the original idea.
Separate process quality from investment outcome
A second reason to preserve the original thesis is that realized returns are an unreliable record of decision quality.
Baron and Hershey demonstrated this in five experiments in which participants evaluated decisions made under uncertainty. Even when participants understood that the original decision maker had the same ex ante information in the good-outcome and bad-outcome versions, they tended to judge the decision or decision maker more favorably when the realized outcome was favorable.
That finding is particularly uncomfortable for investment organizations because the P&L is both indispensable and noisy.
A position can make money for reasons the investor did not anticipate. A strong process can produce a loss when a low-probability adverse event occurs. A weak thesis can survive for years because the relevant falsifying evidence has not arrived. A correct directional forecast can still be a bad investment if it was already embedded in the price.
The answer is not to ignore returns.
It is to evaluate two things separately:
Was the original decision reasonable given what was knowable at the time?
And:
What did subsequently happen?
The living thesis register supplies the first record.
That record is useful at the portfolio-review level as well as for individual investments. CFA Institute Research Foundation’s 2026 work on investment committees argues that unstructured deliberation can weaken accountability when individual inputs are not explicitly recorded and therefore cannot be evaluated later. The monograph proposes a much more formal portfolio-aggregation model than most firms will want to adopt, and part of its analysis relies on simulated committee settings, but the accountability principle is relevant: ex ante judgments need to exist in inspectable form if an organization wants to learn from them.
Without such a record, postmortems tend to become storytelling exercises.
The winner “worked because our thesis was right.”
The loser “failed because conditions changed.”
Both explanations may be true. Neither is a process evaluation.
Run the living-thesis review
The register itself should be deliberately small. If maintaining it requires rewriting the investment memo every week, teams will stop using it.
For each assumption material enough to alter the portfolio decision, record seven fields:
- Assumption: the causal proposition that must hold, stated precisely enough to be wrong.
- Current confidence: the team’s present degree of belief, using a probability or calibrated band appropriate to the evidence.
- Supporting and disconfirming evidence: what has arrived on both sides since the previous review, with provenance where material.
- Leading signpost: the observable most capable of providing information about the assumption before the final outcome.
- Decision threshold: the condition that requires formal reconsideration and the portfolio decisions it could affect.
- Owner: the person accountable for updating the evidence, not for defending the assumption.
- Review date: when the assumption will be reconsidered even if no threshold has fired.
The review meeting should begin with <mark>what changed in the register</mark>, not with a fresh retelling of the thesis.
Which assumptions moved?
What new evidence caused the movement?
Which assumption has accumulated contradictory evidence without a corresponding change in confidence?
Did a threshold fire?
Has the causal chain changed enough that the investment requires re-underwriting?
Which material unknown is receiving no new information and therefore deserves additional research?
The result is not a perfectly Bayesian portfolio manager. Human judgment will still determine how evidence is interpreted, how much weight to place on management behavior or qualitative field research, and when a noisy signal has become meaningful.
But the team gains something valuable: a visible history of its own reasoning.
The investment thesis stops being a document that was written once.
It becomes a record of what the team believed, why it believed it, what evidence arrived, and how that evidence changed, or failed to change, the decision.
Where Enquire fits: preserve the reasoning, not just the latest answer
A living-thesis process creates an information-management problem of its own.
Evidence accumulates across earnings calls, expert conversations, diligence, portfolio reviews, adjacent-company research, regulatory developments, and internal debate. If the original context fragments across notes and meetings, the team eventually knows its latest opinion but loses the path by which it arrived there.
Enquire’s current capital-markets offering is explicitly oriented toward preserving that context. Its capital-markets page describes workflows for pressure-testing investment narratives, identifying assumptions that must hold, tracking how thinking evolves over time, and maintaining continuity across earnings, diligence, and portfolio reviews.
Its current product page also describes an “Evolving Research Context” in which context is preserved across inquiries as new information and expert insights emerge, alongside individual and organization-level Research Centers.
That can support a living-thesis workflow if the technology is used as a memory and research layer rather than as an oracle.
The initial assumptions can provide structure for subsequent research. New expert evidence can be gathered against specific unresolved questions. Contradictory perspectives can be retained rather than collapsed into a single synthesis. Later reviews can revisit earlier analysis with its original context intact.
The important outcome is not that a system determines whether conviction should be 62% rather than 58%.
It is that the team does not have to rely on memory to reconstruct what it once believed and why.
Conviction should be allowed to leave evidence behind
A coherent investment narrative remains valuable.
Analysts need to explain opportunities. Portfolio managers need to make choices. Committees need a compressed account of why an investment belongs in the portfolio.
But compression should come at the end of the reasoning process, not replace it.
The underlying investment thesis should remain decomposed: a set of causal assumptions with identifiable evidence, confidence levels, signposts, and conditions for revision.
That changes what it means to have conviction.
Conviction is no longer the ability to defend the same story through volatility.
It is the willingness to state precisely what you believe, expose that belief to evidence, and revise it without quietly rewriting the past.
A good investment process should therefore be able to answer a question more demanding than “Is the thesis intact?”
What do we believe now that we did not believe when we invested, and what, exactly, changed our mind?
Sources and further reading
- Pavel Atanasov et al., “Small Steps to Accuracy: Incremental Belief Updaters Are Better Forecasters,” Organizational Behavior and Human Decision Processes sciencedirect.com
- Jonathan Baron and John C. Hershey, “Outcome Bias in Decision Evaluation,” Journal of Personality and Social Psychology bear.warrington.ufl.edu
- Barry M. Staw, “Knee-Deep in the Big Muddy: A Study of Escalating Commitment to a Chosen Course of Action” web.mit.edu
- Jeanne W. Ross and Nils O. Fonstad, “Learn from Hypotheses, Not Failures,” MIT CISR cisr.mit.edu
- Chris Bradley, Martin Hirt, and Sven Smit, “How to Confront Uncertainty in Your Strategy,” McKinsey mckinsey.com
- Dan Lovallo and Daniel Kahneman, “Delusions of Success: How Optimism Undermines Executives’ Decisions,” Harvard Business Review store.hbr.org
- CFA Institute, Standard V(A): Diligence and Reasonable Basis cfainstitute.org
- Bernhard Scherer, Investment Committees: Governance and Design Choices, CFA Institute Research Foundation rpc.cfainstitute.org
- Enquire, Capital Markets enquire.ai