Most operational mistakes in fragile environments are not failures of execution. They are failures of timing, registered as failures of execution only after the fact. A team moves an asset, signs a memorandum, deploys a system, or makes a public statement at a moment that, viewed from the next morning, was already past the threshold at which the choice could be revised. The question we are most often asked, by sovereigns and by private principals alike, is what to do. The question we have learned to insist on first is when the doing can still be undone.

This note describes how the firm reasons about that question. It is not a methodology in the proprietary sense. The vocabulary draws on a body of public scholarship in decision theory, organisational sociology, and the operations-research literature on real options.1 What is distinctive is not the framework. It is the discipline of applying it before, rather than after, an engagement is committed.

01.I. THE LAST REVERSIBLE POINT

In any sequence of decisions, there is a moment past which the principal cannot return to the prior state at acceptable cost. Some of these moments are physical: an aircraft that has departed, a transfer that has cleared, a personnel rotation that has been announced. Some are reputational: a public position taken, a partner endorsed, a venue booked. Some are legal: a filing accepted, a jurisdiction selected, a counsel retained. Most are compound, with the irreversible element hidden inside an action that appears procedural.

We refer to the last moment at which the principal retains symmetric optionality as the last reversible point, or LRP. The defining test is not whether a decision can be formally rescinded. Many decisions can be formally rescinded at any time. The test is whether the rescission would leave the principal in substantially the same position they occupied before the decision was taken. If a withdrawal would itself constitute a signal, a liability, or a precedent, the LRP has already passed.

The LRP is not a property of the decision. It is a property of the environment in which the decision is taken. The same action, executed under different conditions, has different irreversibility profiles. Identifying the LRP therefore requires reading the environment, not the action.

02.II. WHY THIS PROBLEM IS HARDER THAN IT LOOKS

The literature on sunk-cost reasoning and escalation of commitment has been mature for four decades.2 In principle, every operator knows that a decision should be evaluated against its forward marginal value, not its accumulated cost. In practice, the forward marginal value of reversal is rarely calculable in real time, and the accumulated cost is always salient. The asymmetry is structural.

Three factors make the LRP harder to see in fragile environments specifically.

The first is signal compression. In stable jurisdictions, irreversible thresholds are typically marked by procedural artefacts: a court date, a closing meeting, a regulatory filing. In fragile environments, the same threshold may be marked by an informal exchange, a meal, or an absence of objection from a counterparty whose silence has standing. The procedural artefact does not arrive, and the threshold is crossed without anyone present recognising the crossing.

The second is counterparty plurality. A decision that is reversible against one counterparty may be irreversible against another. A principal who can withdraw a contractual offer in one capital may have already conceded a position to a regulator, a security service, or a tribal authority observing the same exchange. The LRP must be evaluated against the most consequential observer, not the formal counterparty.

The third is temporal drift. The reversibility of a decision decays continuously, not in steps. There is no alarm that fires at the LRP. The principal experiences a gradient, not a threshold, and gradients are systematically discounted by decision-makers under load.3

03.III. THE TIMESTAMPING DISCIPLINE

The operational response to these three factors is to convert the gradient into a record. Before any consequential action, we require a written estimate of the LRP, expressed as a date and time, with the conditions under which it would advance or retreat. The estimate is not a prediction. It is a forcing function. It compels the analyst to articulate, in advance, what would have to be true for the decision to remain reversible, and at what specific moment those conditions are expected to fail.

The estimate has three components. The first is the anchor event, the observable occurrence that closes the window. The second is the latency band, the range of hours or days within which the anchor event is expected, conditional on current intelligence. The third is the retraction cost, the estimated cost, in resources, reputation, and counterparty trust, of reversing the decision at the LRP minus one hour, the LRP, and the LRP plus one hour. The three figures, taken together, generate the curve against which the principal makes the commit-or-defer call.

The discipline is unforgiving in one respect. If the analyst cannot produce the three components, the engagement is not ready for a commit decision. This is not an aesthetic preference. It is the operational expression of an epistemic claim: that a decision whose reversibility cannot be characterised is a decision that should not be taken under conditions of fragility.

04.IV. WHAT THE DISCIPLINE CHANGES

Three changes follow from applying this method consistently across an engagement.

The first is a different sequencing of work. Tasks that appear urgent often have wide LRP windows, and tasks that appear routine often have narrow ones. A site visit that can be deferred a week has a wider window than a clearance request that cannot be revised after submission. The order of operations is determined by window width, not by perceived priority.

The second is a different posture toward information. Once an LRP has been timestamped, information acquired after that point loses operational value, regardless of its quality. Intelligence is valuable to the extent that it arrives before the window closes. This reframes what the firm spends its collection resources on. We do not pursue the most accurate picture. We pursue the most accurate picture available before the relevant LRP.

The third is a different account of failure. Engagements that go badly are typically described in terms of bad decisions. Under this method, an engagement that goes badly is more often described in terms of a misread LRP, a window that was narrower than the analyst characterised, an anchor event that arrived earlier than the latency band suggested, a retraction cost that climbed faster than the curve predicted. The post-mortem is conducted against the timestamp, not against the outcome. Outcomes are noisy. Timestamps are not.

05.V. LIMITS OF THE FRAMEWORK

Two limits are worth naming.

The framework does not generate the right answer. It generates the right question, on a clock. Whether the principal commits or defers remains a judgement call, made by the principal, against criteria that the framework does not supply. We are explicit with clients about this. The firm does not make decisions on a principal's behalf. It characterises the structure of the decisions the principal is about to make.

The framework also degrades under conditions of strategic ambiguity. When a counterparty's intentions are deliberately opaque, the LRP cannot be estimated within usable bounds. In those cases, the correct application of the method is to refuse the commit-or-defer framing entirely and to substitute a different one, in which the engagement is reduced to a sequence of small, individually reversible moves, each timestamped against its own anchor. This is slower. It is also the only honest posture.

06.VI. WHY THIS MATTERS NOW

The volume of cross-jurisdictional engagement has increased faster than the institutional capacity of most principals to characterise irreversibility. Counsel, in most cases, is delivered against a frame in which decisions are described as reversible until they are described as wrong. The interval between those two descriptions is where most of the cost in a fragile engagement is realised. Closing that interval is the substantive work of an advisory firm operating in this environment. The analyst who can name the LRP, and timestamp it, is the analyst whose counsel survives the morning after.

The firm's standing practice is to attach an LRP estimate to every engagement memorandum, before commit. Where the estimate cannot be produced, the engagement is held. Where it can, the timestamp is the artefact against which the engagement is later reviewed. We do not claim that this method eliminates error. We claim that it relocates error from the moment of the outcome to the moment of the analysis, where it can be examined, contested, and corrected before the principal has paid for it.

07.Footnotes

  1. For the foundational treatment of real options under irreversibility, see Avinash K. Dixit and Robert S. Pindyck, Investment under Uncertainty (Princeton University Press, 1994), particularly chapters 2 and 5 on the value of waiting and the asymmetric cost of reversal. The argument applies, with appropriate modification, to non-financial decisions taken under conditions of partial information. ↩

  2. Barry M. Staw, "The Escalation of Commitment to a Course of Action," Academy of Management Review 6, no. 4 (1981): 577-587, remains the canonical statement. For the organisational extension, see Jerry Ross and Barry M. Staw, "Organizational Escalation and Exit: Lessons from the Shoreham Nuclear Power Plant," Academy of Management Journal 36, no. 4 (1993): 701-732. ↩

  3. On temporal discounting and decision-maker load, see Daniel Kahneman, Thinking, Fast and Slow (Farrar, Straus and Giroux, 2011), and the operational extension in Gary Klein, Sources of Power: How People Make Decisions (MIT Press, 1998), particularly the chapters on recognition-primed decision under time pressure. ↩