Leadership | Strategy | Decision-Making

The leader’s hardest decision is often not what to start, but what to stop. When evidence changes, disciplined reversal can preserve capital, talent and trust. Persistence after the strategic thesis has failed does the opposite.

By Frank Farnel | Responsible Public Affairs | September 29, 2026

Executive Summary

• Organizations are structurally better at launching initiatives than ending them. Budgets, careers, public commitments and identity accumulate around a strategy, making reversal politically harder even when its economics weaken.
• Research on escalation of commitment shows why decision-makers may continue investing after negative evidence appears. Sunk costs are not only financial; leaders may also defend their reputation, prior forecasts and organizational narrative.
• A credible exit separates the enduring objective from the current vehicle for achieving it. Ford’s EV reset, NASA’s redesign of Mars Sample Return, Google’s closure of Stadia and Microsoft’s retreat from Nokia phone hardware illustrate very different versions of that choice.
• The Strategic Exit Test examines five questions: which premise failed, whether the evidence is decisive, which option should replace simple persistence, what obligations the organization owes affected stakeholders, and how resources will be redeployed.
• Leaders should define review dates and stop criteria before enthusiasm and sunk costs distort judgment. They should also separate the decision reviewer from the executive whose identity is tied to the original commitment.

The Launch Bias

Leadership literature has an affection for beginnings. We celebrate the new strategy, the transformation program, the acquisition, the product launch and the bold commitment. Beginnings create photographs, speeches and promotions. They allow leaders to describe uncertainty as ambition.

Endings are different. They create an immediate accounting charge, disappointed employees, stranded suppliers and difficult questions about the judgment that authorized the initiative. A canceled project is visible. The capital that might have been destroyed by keeping it alive is hypothetical. That asymmetry makes persistence look responsible longer than it should.

The relevant distinction is not between leaders who remain committed and leaders who change their minds. Every serious strategy requires commitment. Capabilities take time to build, markets fluctuate and early performance is often noisy. An organization that abandons every project after a disappointing quarter is not adaptive; it is ungovernable.

The real leadership test is whether commitment remains attached to an objective or becomes attached to a specific plan. A company may still believe in electrification while canceling vehicles that cannot earn an acceptable return. A space agency may still value Martian samples while rejecting an unaffordable mission architecture. A technology group may retain a platform’s engineering while closing the consumer service built on top of it.

A strategy should be defended by evidence about the future, not by the size of the investment already made in the past.

That sentence sounds obvious. In practice, organizations regularly reverse it.

Why Smart Organizations Continue Too Long

Barry Staw’s foundational 1976 experiment described escalation of commitment: the tendency to allocate further resources to a chosen course of action after receiving negative feedback, particularly when the decision-maker was responsible for the original choice.^[11] Hal Arkes and Catherine Blumer later documented the sunk-cost effect—the greater tendency to continue an endeavor once money, effort or time has already been invested.^[12]

These findings do not mean every continuation decision is irrational. Negative signals can be temporary, abandoning a project can destroy recoverable value, and a rival may benefit if an organization withdraws too soon. The more useful point is that prior investment can contaminate a decision that should be made on forward-looking grounds.

A 2012 meta-analysis by Dustin Sleesman and colleagues organized the drivers of escalation across project, psychological, social and structural factors.^[13] That breadth matters. Persistence is rarely caused by one stubborn executive alone. It can be reinforced by incentive systems, fragmented information, public promises, completion targets, group loyalty and the absence of an attractive alternative.

Five pressures are especially relevant to senior leaders.

1. Personal authorship. A leader who sponsored a project may hear “stop” as “admit incompetence.” The organization then evaluates the executive’s identity rather than the project’s future cash flows or public value.
2. Near-completion logic. Projects that are “almost finished” can remain almost finished through repeated budget cycles. Completion percentage is persuasive even when the remaining technical risk contains most of the uncertainty.
3. Narrative lock-in. Once a strategy has been presented to investors, employees or governments as evidence of courage, reversal appears to threaten credibility. Leaders may continue in order to protect the story they told.
4. Distributed harm. The cost of continuation may be spread across many budgets and years, while the cost of stopping arrives immediately in one impairment, redundancy program or political controversy.
5. No designed exit. If contracts, technology and teams were built on the assumption of uninterrupted continuation, stopping destroys more value than necessary. The lack of an exit option becomes an argument for further commitment.

Real-options theory offers a useful counterweight. Under uncertainty, management creates value not only by choosing the right investment, but also by preserving the ability to expand, delay, contract or abandon as information improves.^[15] Flexibility is not indecision. It is an asset—provided the organization defines when and how it will be used.

The Strategic Exit Test

The Strategic Exit Test is a leadership framework for deciding whether to continue, narrow, pause, partner, sell or stop an initiative. It is deliberately broader than a financial hurdle rate because an exit affects customers, employees, regulators, suppliers and institutional credibility.

1. Premise: What Had to Be True?

Every strategy rests on a small number of decisive assumptions: demand will reach a given level; a technology will meet a performance threshold; regulation will remain favorable; unit economics will improve with scale; a partner will deliver; the organization can integrate the acquisition.

Leaders should restate those assumptions in present tense. Which remain true? Which have weakened? Which were never tested? This prevents a strategy from surviving merely because its original language was broad enough to absorb any outcome.

2. Evidence: Is the Signal Strong Enough?

One poor quarter is not a failed thesis. Neither is one enthusiastic forecast proof of recovery. The decision requires a defined evidence threshold: repeated customer behavior, technical test results, independent cost estimates, regulatory change or a sustained gap between actual and required economics.

Evidence should be presented in a form that makes opportunity cost visible. The question is not only “Can this project still work?” Almost any project can work under sufficiently generous assumptions. The more demanding question is “Is this still the best use of the next dollar, the next engineer and the next year?”

3. Options: Is the Choice Really Continue or Kill?

Binary framing creates unnecessary resistance. A program can be narrowed to its strongest market, staged behind a technical milestone, transferred to a partner, separated from the parent, licensed, sold or converted into infrastructure used elsewhere. Full continuation and total liquidation are only two points on a wider decision map.

4. Obligations: Who Pays for the Exit?

A strategy can fail commercially while the organization retains obligations to people who relied on it. Employees may have specialized their careers. Customers may have purchased hardware or stored data. Governments may have funded infrastructure. Suppliers may have expanded capacity. Responsible exit design identifies those dependencies before the announcement.

This is not philanthropy. The treatment of stakeholders during one closure changes the credibility of the next launch. Customers remember whether they were refunded, trapped or ignored. Employees observe whether bad news was hidden until the last possible moment.

5. Reallocation: What Moves, and Where?

A good exit releases more than cash. It can recover technology, data, intellectual property, contracts, facilities and experienced teams. It can also reveal which assumptions and governance processes failed. If those assets remain stranded, the organization has paid twice: once for the failed thesis and again for refusing to learn from it.

Four Cases, Four Kinds of Exit

Ford: Stopping Scale Without Abandoning the Direction

Verified facts. In December 2025, Ford announced a major reset of its electric-vehicle roadmap. It said it would discontinue the current-generation F-150 Lightning, cancel three planned EVs and redirect investment toward hybrids, extended-range electric vehicles, a lower-cost universal EV platform and battery energy-storage systems. Ford expected approximately $19.5 billion in special items, most of them associated with asset write-downs and changes to its battery joint venture.^[1] A related SEC filing identified about $8.5 billion of expected pre-tax write-downs tied to EV manufacturing capacity and canceled products.^[2]

The company did not present the decision as an exit from electrification. In Europe, it simultaneously pursued a partnership with Renault Group to use the Ampere platform and manufacturing system for two new Ford-branded electric cars. In its second-quarter 2026 update, Ford said its Model e unit had posted a third consecutive quarter of year-over-year profit improvement, while acknowledging substantial charges linked to the cancellations and the BlueOval SK transaction.^[3]

Analysis. Ford illustrates a selective strategic exit. The company separated a directional belief—electrified transport—from particular products, capacity assumptions and ownership choices. It stopped treating every earlier EV commitment as one indivisible strategy.

The scale of the charge also shows the cost of waiting until assets and programs have accumulated. Yet a large impairment does not by itself prove that the reset was correct or late. That judgment depends on the returns generated by the replacement portfolio, the evolution of EV demand and regulation, and whether Ford can compete in regions where electrification continues at a different pace.

Evidence limitation. Ford’s improved quarterly comparison is an early operating indicator, not proof that the 2025 reset will create long-term value. The company’s 2029 profitability objective remains forward-looking.

NASA: Preserve the Mission, Replace the Architecture

Verified facts. Mars Sample Return is intended to bring scientifically selected material collected by the Perseverance rover back to Earth. An independent review and subsequent NASA analysis concluded that the then-current architecture would probably cost between $8 billion and $11 billion and could return samples in 2040. In April 2024, NASA Administrator Bill Nelson publicly said an $11 billion budget was too expensive and a 2040 return date too distant; the agency sought revised designs with lower cost, less complexity and earlier delivery.^[4]

NASA then commissioned studies across its centers and industry. In January 2025, it announced that it would continue examining two landing approaches before confirming a new architecture, cost and schedule.^[5] The NASA Office of Inspector General had separately recommended that the program stabilize design, create realistic life-cycle estimates and consider launch scenarios before advancing.^[6]

Analysis. This is strategic reversal without abandonment of purpose. NASA did not say the scientific objective had lost value. It acknowledged that the vehicle chosen to deliver that objective no longer met affordability and schedule constraints.

The distinction is important in mission-driven organizations. Teams often defend a plan by invoking the nobility of the goal. But commitment to the goal can require rejecting an architecture that consumes the resources needed for other missions. NASA’s public use of independent review, alternative studies and explicit cost-and-date boundaries created institutional permission to revisit the plan without pretending the scientific ambition had disappeared.

Evidence limitation. The redesign process demonstrates a disciplined reset mechanism, not a completed success. Final performance can be judged only after an architecture, budget and schedule are approved and executed.

Google Stadia: Make the Exit Part of the Product Experience

Verified facts. Google launched Stadia as a consumer cloud-gaming service in 2019. On September 29, 2022, it announced that the service had not gained the expected user traction and would close. Users retained access until January 18, 2023. Google offered refunds for Stadia hardware purchased through the Google Store and for games and add-on content purchased through the Stadia Store, while excluding most prior Stadia Pro subscription fees.^[7]

Google also described plans to apply the underlying streaming technology elsewhere and said many team members would move to other parts of the company. Its shutdown FAQ documented refund mechanics, access dates, hardware treatment and limits on transferring game progress.^[8]

Analysis. Stadia is often cited simply as a failed Google product. That is accurate at the consumer-service level but incomplete as an exit analysis. The service closed; some technology and talent were retained; customers received a defined wind-down period and unusually broad refunds.

The case shows that leadership continues after the stop decision. A product’s final user experience is its closure. Refund policy, data portability, support, partner communication and the timing of service termination determine whether the organization converts disappointment into hostility.

Stadia also carries a warning. Google had already closed its internal Stadia game-development studios in February 2021, well before announcing the consumer service shutdown. That earlier contraction did not make closure inevitable, but it was a material signal. Leaders should ask whether a series of “narrow adjustments” is really evidence that the original integrated thesis has dissolved.

Microsoft and Nokia: The Cost of a Late Admission

Verified facts. Microsoft completed its acquisition of substantially all of Nokia’s Devices and Services business on April 25, 2014. Its 2015 annual report placed the purchase price at $9.4 billion, including $1.5 billion in cash acquired.^[10] In July 2015, Microsoft announced a restructuring of its phone hardware business and expected an impairment charge of approximately $7.6 billion plus $750 million to $850 million in restructuring charges.^[9] The final fiscal-year results included $7.5 billion of goodwill and asset impairment charges and $780 million of restructuring expense associated with phone hardware.

Analysis. The speed and scale of the write-down make this a classic late-exit case. The acquisition had sought to strengthen Microsoft’s position in mobile devices. By the following fiscal year, the company had concluded that the standalone phone strategy could not justify the acquired asset value.

The lesson is not that leaders should never make large acquisitions or that every write-down reveals negligence. Acquisitions are made under uncertainty, and market structure can shift. The sharper lesson is that integration plans need thesis-specific review points from the start. If the deal depends on device share, developer adoption, distribution leverage or ecosystem effects, leaders must define how much evidence and time each proposition receives before capital is redirected.

A late admission can still be better than indefinite escalation. But it leaves less value to recover, fewer stakeholder options and a larger credibility gap between the original promise and the final accounting.

Comparison: What Each Exit Preserved

Case	What changed	Exit form	Value preserved	Leadership risk
Ford EV reset	Product economics, policy and demand assumptions	Cancel, narrow, partner and redeploy	Affordable-EV platform, European partnership, hybrid and storage capabilities	Confusing a U.S. reset with a universal view of global EV demand
NASA Mars Sample Return	Affordability and schedule of the mission architecture	Independent review and redesign	Scientific objective, cached samples and international cooperation	Allowing review to become indefinite delay without a final executable plan
Google Stadia	Expected consumer traction	Orderly service closure and refunds	Streaming technology, talent and some partner applications	Underestimating the ecosystem trust lost when purchased digital access ends
Microsoft/Nokia phones	Strategic value of the acquired hardware business	Impairment and restructuring	Selected mobile software, patents and broader cloud strategy	Recognizing thesis failure only after integration and asset value had deteriorated

Practical Lessons

First, distinguish the destination from the vehicle. An enduring ambition should not immunize one program from review. Electrification is not one vehicle line. Scientific discovery is not one mission architecture. Participation in mobile computing was not identical to owning a handset business.

Second, write the stop criteria while optimism is still high. A proposal should specify the assumptions that justify it, the dates on which they will be reviewed, the evidence that would trigger redesign and the authority permitted to stop funding. This is easier before careers and public narratives attach themselves to the plan.

Third, separate review from authorship. The original sponsor should explain the strategy but should not control the entire evaluation of its continuation. Independent technical review, a separate capital committee or an external assurance process can reduce self-justification without stripping leaders of accountability.

Fourth, price opportunity cost. A struggling initiative can look recoverable when evaluated in isolation. Compare it with the best alternative use of the same capital, talent and management attention. Continuing must defeat that alternative, not merely produce a plausible upside scenario.

Fifth, design the stakeholder exit before it is needed. Contracts should address data transfer, customer refunds, supplier ramp-down, employee redeployment and intellectual-property ownership. A reversible strategy is cheaper to change because its dependencies were not hidden.

Sixth, make learning transferable. The post-exit review should distinguish a bad objective, a bad assumption, a bad execution and a bad governance process. Those are not the same failure. If they are collapsed into one generic “lesson learned,” the next strategy will repeat the same error under different terminology.

What Leaders Should Do Now

1. Select three major commitments for an exit review. Choose initiatives with high capital use, long duration, strategic visibility or stakeholder dependence. Do not begin with small, politically easy projects.
2. Reconstruct the original thesis. List the five or fewer assumptions that made the initiative rational when approved. Record which have been validated, weakened or disproved.
3. Commission a forward-only case. Ask what the organization would invest from today if it did not already own the asset, employ the team or make the public promise. The answer does not decide the outcome, but it exposes the weight of sunk costs.
4. Demand at least four options. Compare full continuation, narrower scope, partnership or sale, and orderly closure. Include timing, cash impact, capability retention and stakeholder treatment.
5. Name an independent decision owner. Give someone without reputational dependence on the original choice the authority and information to challenge it. Independence should be procedural, not theatrical.
6. Prepare the transition before the announcement. Determine who is told first, which customer or employee obligations apply, what data and technology move, which supplier commitments remain and where released resources go.
7. Explain the change without rewriting history. State what the organization believed, what evidence changed, what remains valid, what is ending and what happens next. Avoid claiming that the discontinued plan was secretly a success or that no mistake occurred.
8. Set a review date for the replacement. A reset can itself become a new orthodoxy. Apply the same discipline to the successor strategy.

Conclusion: Persistence Is Not the Same as Conviction

Leaders earn credibility by keeping promises, but not by preserving every plan unchanged. A promise to protect the organization’s future, serve customers or deliver public value is larger than the initiative chosen at one moment to fulfill it.

The strongest strategic exits are neither impulsive retreats nor public-relations exercises. They identify which premise has failed, use evidence that can withstand challenge, preserve useful capabilities, meet obligations to people who relied on the strategy and move resources toward a better proposition.

There is no leadership glory in canceling for its own sake. Constant reversal exhausts organizations and destroys trust. Yet there is also no virtue in asking employees, investors, taxpayers or customers to finance a strategy because admitting its expiration would be uncomfortable.

A company can survive a failed thesis. It is less likely to survive a leadership system that cannot recognize one.

Key Evidence

• $19.5 billion: Ford’s expected special items associated with its December 2025 EV reset, with most recognized in the fourth quarter of 2025. Source [1].
• Three planned EVs canceled: Ford also ended production of the current-generation F-150 Lightning while retaining other electrification investments. Source [2].
• $8 billion–$11 billion and 2040: the likely cost range and return date associated with the Mars Sample Return architecture NASA rejected as unaffordable and too slow. Source [4].
• January 18, 2023: the Stadia shutdown date; Google offered refunds for eligible hardware, games and add-ons and provided a defined transition period. Source [7].
• $9.4 billion acquisition; $7.5 billion impairment: Microsoft’s reported Nokia Devices and Services purchase price and the phone-hardware impairment recorded in fiscal 2015. Source [10].

Glossary

Escalation of commitment: continued investment in a chosen course of action despite negative feedback, often reinforced by responsibility for the original decision.

Impairment: an accounting reduction in an asset’s carrying value when the organization concludes that the asset will not recover its recorded value.

Opportunity cost: the value of the best alternative forgone when resources are committed to a particular choice.

Real option: managerial flexibility to expand, delay, contract, switch or abandon an investment as uncertainty resolves.

Sunk cost: money, time or effort already spent and not recoverable; it should not determine a forward-looking choice merely because it was incurred.

Strategic exit: a governed decision to stop, narrow, transfer or replace a strategy or initiative while managing obligations and recovering useful value.

References and Further Reading

Official and Primary Case Sources

1. Ford Motor Company. Ford Follows Customers to Drive Profitable Growth; Reinvests in Trucks, Hybrids, Affordable EVs, Battery Storage; Takes EV-Related Charges. December 15, 2025.
2. Ford Motor Company. Current Report: EV Manufacturing Capacity and Product Roadmap Decisions. U.S. Securities and Exchange Commission, December 2025.
3. Ford Motor Company. Four Things to Know About Ford’s Second-Quarter 2026 Results. July 2026.
4. NASA. NASA Sets Path to Return Mars Samples, Seeks Innovative Designs. April 15, 2024.
5. NASA. NASA to Explore Two Landing Options for Returning Samples from Mars. January 7, 2025.
6. NASA Office of Inspector General. Audit of the Mars Sample Return Program. Report IG-24-008, February 28, 2024.
7. Phil Harrison, Google. A Message About Stadia and Our Long-Term Streaming Strategy. September 29, 2022.
8. Google. Stadia Announcement FAQ. Last updated March 8, 2023.
9. Microsoft. Microsoft Announces Restructuring of Phone Hardware Business. July 8, 2015.
10. Microsoft. Microsoft 2015 Annual Report. 2015.

Academic and Theoretical Sources

11. Barry M. Staw. Knee-Deep in the Big Muddy: A Study of Escalating Commitment to a Chosen Course of Action. Organizational Behavior and Human Performance, Vol. 16, No. 1, June 1976, pp. 27–44.
12. Hal R. Arkes and Catherine Blumer. The Psychology of Sunk Cost. Organizational Behavior and Human Decision Processes, Vol. 35, No. 1, 1985, pp. 124–140.
13. Dustin J. Sleesman, Donald E. Conlon, Gerry McNamara and Jonathan E. Miles. Cleaning Up the Big Muddy: A Meta-Analytic Review of the Determinants of Escalation of Commitment. Academy of Management Journal, Vol. 55, No. 3, 2012, pp. 541–562.
14. Ramiro Montealegre and Mark Keil. De-Escalating Information Technology Projects: Lessons from the Denver International Airport. MIS Quarterly, Vol. 24, No. 3, September 2000, pp. 417–447.
15. Lenos Trigeorgis and Jeffrey J. Reuer. Real Options Theory in Strategic Management. Strategic Management Journal, Vol. 38, No. 1, 2017, pp. 42–63.

Source and Methodology Note

Research was completed on September 29, 2026. The article prioritizes company filings, official corporate disclosures, NASA and NASA Office of Inspector General materials, and peer-reviewed research. Corporate statements establish what an organization announced, recorded or said it intended; they are not independent proof that the resulting strategy succeeded.

The Ford reset and NASA redesign remain works in progress. Their inclusion illustrates decision mechanisms and trade-offs, not final performance. Google’s refund and closure terms are verified from company sources, but the value ultimately recovered from Stadia technology and talent is not quantified here. Microsoft’s acquisition and impairment figures are historical accounting facts; the interpretation that the case represents a late strategic exit is analysis. The Strategic Exit Test is the author’s synthesis of the cited research and cases, not a validated diagnostic instrument.

Suggested Internal Links

• The Last Five Percent Is Not Waste: The Leadership Case for Strategic Slack — for the complementary question of which options and reserves should be preserved before a reversal becomes necessary.
• The Apology Is Not the Repair: What Leaders Must Rebuild After Institutional Failure — for the responsibilities that follow when delayed action has already caused harm.
• The Face on the Screen Is No Longer Proof — for a related discussion of decision authority, verification and the right to challenge urgency.

Hashtags: #Leadership #StrategicDecisionMaking #CorporateStrategy