Innovation portfolios rarely fail because leaders stop too many initiatives too early. They fail because organizations keep funding weak initiatives after the original assumptions have broken.
The longer a project runs, the harder it becomes to challenge. Budgets have been approved. Teams have been formed. Executives have defended the decision. Public commitments may have been made. Progress activity creates the appearance of momentum even when the evidence is deteriorating.
The result is commitment escalation: more time, money, and reputation are invested because stopping feels like admitting failure.
Mature leadership requires the opposite capability. It requires **kill criteria**—pre-agreed evidence thresholds that determine when an initiative should continue, change, pause, transfer, or stop.
The purpose is not to punish experimentation. It is to protect learning, capital, attention, and institutional credibility from projects that survive through inertia.
Vietnam's strategic-technology program signals a more disciplined model
On September 28, Government News reported on a new financial mechanism for strategic-technology research and development. The mechanism links funding to work packages, evaluation milestones, and outcomes; allows multiple research approaches; recognizes acceptable research risk; and requires resources to be moved away from tasks that do not meet agreed milestones.
The review found recurring weaknesses: duplicated proposals, unclear separation between investment and research costs, missing adoption units, and the absence of explicit milestones for deciding whether a project should continue or stop.
By September 27, ministries and agencies had proposed 38 tasks with total funding of VND 13.75 trillion. Thirty had been appraised, 22 were eligible, 14 approved, eight not eligible, and eight still awaiting appraisal. The Ministry of Science and Technology explicitly proposed that unready tasks should not proceed and that disbursement should not become the objective.
That principle is relevant far beyond public R&D.
Every transformation portfolio—AI, digital, product, market expansion, education reform, or organizational change—needs a legitimate mechanism for stopping work that no longer deserves scarce resources.
What kill criteria mean
**Kill criteria are explicit, evidence-based conditions agreed before or during an initiative that trigger a decision to stop, pause, redesign, transfer, or materially reduce investment.**
The word “kill” can sound severe, but the concept is not anti-innovation.
Innovation contains uncertainty. A disciplined portfolio expects some assumptions to fail. Kill criteria allow the organization to discover that failure at lower cost, preserve the learning, and redirect resources toward stronger opportunities.
Without kill criteria, decisions depend on political influence, executive mood, budget timing, or exhaustion. The weakest projects may survive because their sponsors are powerful, while promising experiments lose resources because they are newer or less visible.
Why leaders escalate commitment
Sunk cost becomes an argument
Teams say, “We have already invested too much to stop.” But past expenditure cannot improve future value. The relevant question is whether the next unit of time, money, or attention has a stronger expected return here than elsewhere.
Sunk cost becomes dangerous when it is combined with incomplete evidence. Leaders keep investing not because the project is justified, but because the organization has no accepted way to close it.
Sponsorship becomes identity
Senior leaders often attach personal credibility to major initiatives. Once a project becomes “the CEO's program” or “the ministry's flagship,” evidence is interpreted defensively.
Problems are described as implementation delays. Missed milestones are reframed as ambition. Additional budget is presented as the last step required to unlock value.
Healthy sponsorship gives a project authority. Unhealthy sponsorship makes the project impossible to challenge.
Activity substitutes for progress
Innovation projects generate visible work: meetings, prototypes, hiring, partnerships, procurement, reports, and demonstrations. These activities make the initiative feel alive.
But activity does not prove that the underlying uncertainty is being reduced.
A portfolio review should ask what has been learned about technical feasibility, user behavior, adoption, economics, risk, and strategic value. If the answer is unclear, the project may be producing motion rather than evidence.
Teams hide bad news because stopping is punished
If project closure damages careers, teams will delay or soften negative information. They will protect the initiative until the evidence becomes impossible to ignore.
The organization then learns late and expensively.
Leaders who want honest evidence must make disciplined stopping respectable. A team that invalidates a critical assumption early may have created more value than a team that delivers an unusable system on schedule.
Build kill criteria before momentum takes over
1. State the investment thesis
Every initiative should begin with a concise answer to five questions:
- What strategic problem are we solving?
- What must become true for this investment to create value?
- Which assumptions are most uncertain?
- What evidence would increase or decrease confidence?
- What alternative uses exist for the same resources?
The thesis gives the review process a reference point. Without it, teams can continuously redefine success.
2. Separate failure types
Not every weak result should trigger the same action.
A project may fail because:
- the problem is not important enough;
- the solution is technically infeasible;
- users will not adopt it;
- economics are unfavorable;
- required data or infrastructure is unavailable;
- regulation blocks deployment;
- the organization lacks an adoption owner;
- a better external solution already exists;
- another project has made it redundant.
Some failures justify redesign. Others justify transfer, procurement, or closure. Precision prevents leaders from treating every problem as a request for more time.
3. Define leading and terminal criteria
**Leading criteria** show whether the initiative is learning and moving toward viability: data access secured, critical experiment completed, user behavior observed, integration risk reduced, or adoption owner confirmed.
**Terminal criteria** describe conditions under which continuation is no longer justified: repeated failure of a core assumption, no credible recipient, economics outside the acceptable range, risk beyond tolerance, or duplication with a stronger initiative.
Both matter. Leading criteria prevent premature closure. Terminal criteria prevent indefinite rescue.
4. Set milestone decisions, not milestone presentations
Many stage reviews are designed to communicate progress, not make decisions. Slides are polished, risks are acknowledged, and the project continues by default.
A real milestone review must choose among defined actions:
- continue at the current level;
- continue with conditions;
- redesign the thesis or scope;
- pause until a dependency is resolved;
- transfer to another owner;
- buy or partner instead of build;
- stop and preserve learning.
No decision should mean no automatic funding.
This strengthens [closed-loop execution](/blog/leaders-need-closed-loop-execution): a directive becomes management only when evidence returns and changes the next allocation decision.
5. Protect independent challenge
The project sponsor should not be the only person deciding whether the project survives.
Reviews should include people who understand the domain, technology, economics, adoption context, and portfolio tradeoffs. The challenge function must be able to ask whether the project is still the best route to the outcome.
For high-stakes initiatives, the organization may use a red team, external technical review, user council, or investment committee. Independence matters most when reputation and sunk cost are highest.
6. Design the shutdown pathway
Projects continue partly because stopping creates operational complexity.
A shutdown plan should cover:
- team reassignment;
- contractual obligations;
- data and intellectual-property preservation;
- customer or partner communication;
- reusable code, equipment, or research;
- security and compliance closure;
- documentation of assumptions and evidence;
- decisions about future reconsideration.
Stopping well is an operating capability. Without it, the organization pays a “closure tax” that encourages avoidance.
Manage the portfolio, not isolated projects
Kill criteria create the most value when connected to portfolio allocation.
The central question is not only whether a project is acceptable. It is whether it is stronger than the next-best use of resources.
A portfolio board should compare initiatives across:
- strategic importance;
- evidence strength;
- uncertainty reduction;
- time to meaningful outcome;
- adoption readiness;
- capability created;
- downside risk;
- option value;
- resource concentration;
- duplication and dependency.
This prevents a mediocre legacy initiative from consuming resources that could support a smaller but more promising experiment.
It also complements [risk-weighted attention](/blog/leaders-need-risk-weighted-attention). Leaders should not oversee every project equally; they should focus attention where consequence, uncertainty, and resource exposure are highest.
Distinguish a productive stop from a failed project
A productive stop produces reusable value.
The team should leave behind:
- validated and invalidated assumptions;
- experiment results;
- user and operational evidence;
- architecture and integration lessons;
- cost and timeline reality;
- regulatory findings;
- transferable assets;
- recommendation for future conditions that might justify reopening.
This is how a portfolio compounds learning even when individual initiatives do not scale.
The failure is not that an uncertain project stopped. The failure is that the organization spent heavily without learning what should change.
The leadership behavior behind disciplined stopping
Kill criteria will not work if leaders celebrate launches but hide closures.
Leaders must model four behaviors.
Separate ego from evidence
A leader's responsibility is not to prove the original decision was correct. It is to improve the next decision as evidence changes.
Reward early truth
Teams should gain credibility when they surface a broken assumption with strong evidence. This encourages honest reporting before problems become expensive.
Make alternatives visible
Stopping becomes easier when the organization can see where people and funding will go next. Reallocation turns closure from loss into strategic movement.
Preserve ownership
Accountability does not disappear when risk is accepted. As [value-capture governance](/blog/leaders-value-capture-governance) argues, investment must still be connected to measurable capability and value. Leaders should own the quality of the thesis, milestone design, evidence, and allocation decision.
A simple kill-criteria template
For each major initiative, document:
- strategic outcome;
- critical assumptions;
- evidence required at each stage;
- acceptable cost and time range;
- leading indicators;
- terminal conditions;
- decision authority;
- independent reviewers;
- closure pathway;
- reallocation destination.
Review the criteria when material context changes, but do not rewrite them simply to protect the project.
Conclusion
Innovation requires persistence, but persistence is not the same as escalation.
The mature organization does not stop at the first difficulty. It also does not continue merely because money, identity, and public commitment have accumulated.
Kill criteria create a legitimate middle ground: disciplined experimentation, honest evidence, explicit decisions, and timely reallocation.
Leaders build stronger portfolios when they make it safe to discover that an idea should stop—and difficult for weak initiatives to survive without proof.
Key Takeaways
- Commitment escalation keeps weak initiatives alive after key assumptions fail.
- Kill criteria define evidence-based conditions for continuing, redesigning, pausing, transferring, or stopping.
- Milestone reviews must make allocation decisions, not merely present progress.
- Productive closure preserves learning, assets, and talent for stronger opportunities.
- Leaders must reward early truth and separate personal sponsorship from project evidence.
FAQ
What are kill criteria in innovation management?
Kill criteria are explicit, evidence-based conditions that trigger a decision to stop, pause, redesign, transfer, or materially reduce investment in an initiative.
Do kill criteria discourage innovation?
No. They make experimentation safer by limiting downside, rewarding early learning, and releasing resources from weak initiatives so stronger opportunities can be funded.
When should a leader stop an innovation project?
A leader should consider stopping when a critical assumption repeatedly fails, no credible adoption owner exists, economics or risk exceed agreed limits, a better external solution exists, or the initiative duplicates stronger work.
How should teams document a stopped project?
They should preserve assumptions, experiments, evidence, user learning, technical findings, costs, reusable assets, closure obligations, and conditions that might justify future reconsideration.