Fund, fix or stop

A value creation plan (VCP) commits capital to growth bets whose commercial case was usually a projection at the time of the deal: a new product line, entry into a new vertical, cross-selling after an add-on acquisition, or an AI program meant to lift margin or revenue. Some of them work. Some stall, and the board has to decide whether to fund the next tranche.

That decision is harder than it looks, for a structural reason. The main input is management's forecast, and management has to believe in the plan in order to run it. The deal team that underwrote the bet has its own reasons to see it succeed. The board meets quarterly and sees status reports that measure activity against the plan, which says whether the team is executing but not whether the plan was right.

The cost of getting this wrong has gone up. Bain's 2026 private equity report puts buyout holding periods at around seven years and estimates that a 2.5x return over five years now needs 10% to 12% annual EBITDA growth [1]. A program that absorbs two more years of capital on a wrong premise takes a large share of the hold with it.

This paper sets out the four decisions a board can take on a stalled growth bet, the evidence each one needs, how to write a Reset so that it tests something new instead of buying time, how to stop without turning the decision into a verdict on people, and how to tie the next tranche to evidence gates the board can check. An appendix lists the questions to ask before the next funding vote.

Initiatives in a VCP tend to stall in a few recognizable ways.

A new product line meets its launch dates and misses its bookings. An entry into an adjacent vertical wins early logos and then stops expanding. An add-on acquisition was supposed to open cross-selling into the platform's customer base, and the customers buy one product or the other but rarely both. An AI program delivered working pilots at two sites and has not reached a third.

In each case the plan made a commercial claim about buyers: that they have the problem, will pay for this solution, can be reached through this channel, and will expand at this rate. When the numbers fall short, two explanations compete. Either the premise was wrong, or the premise was right and the execution was weak. These call for opposite responses.

More execution rescues a sound premise and deepens the loss on a wrong one.

The difficulty of telling the two apart from inside the company is covered in Too successful to stop, too unproven to scale. This paper takes the board's side of the same problem.

A board deciding on the next tranche usually has three inputs: the forecast, a status report against milestones, and management's explanation of the gap. All three come from the team running the program.

That is not a criticism of management. A leadership team has to commit to the plan to make it work, and a team that openly doubted its own plan would not execute it well. The same commitment makes management a poor source of an independent view on whether the plan's premise holds.

The deal team is in a similar position. The program was part of the investment case. Barry Staw's research on escalating commitment found that people allocated the most additional resources to a failing course of action when they were personally responsible for the original decision [2]. The effect does not depend on bad faith. It is how people respond when a decision they made is not working.

Board rhythm adds a third problem. A quarterly meeting reviewing a monthly report sees a trend late, and a single bad quarter can always be explained. By the time three quarters confirm a pattern, a large part of the next tranche is already committed.

The practical consequence is that the funding decision needs one input that does not come from inside the program: evidence from the buyers whose behavior the plan depends on, gathered and graded by someone with no stake in the answer.

A board has four real options on a stalled program. Each needs a different kind of evidence, and each should be written down as a decision with its reasoning, not left as a default.

Decision	What has to be true	Evidence the board should see
Fund as planned (Proceed)	The buyer premise holds. The shortfall is execution: sales capacity, implementation, timing.	Scored buyer interviews confirm urgency, budget and authority at the planned price. The gap is explained by named execution causes with a fix and a date.
Change the approach (Pivot)	The problem is real and the buyers exist, and one part of the approach is wrong: price, segment, channel or packaging.	Buyer evidence identifies which dimension fails and what the buyers would accept instead. The revised plan changes that dimension and keeps the rest.
Reframe and retest (Reset)	The evidence is not strong enough either way. The original premise cannot be confirmed or rejected as written.	A new hypothesis, stated so it can fail, with pass/fail thresholds, a capped budget to test it, and a date for the next decision.
Stop	The buyer premise is disproven: the buyers do not have the problem at this priority, do not hold the budget, or will not pay a price that works.	Scored evidence against thresholds declared in advance, and a documented record of what was tested and why it failed.

The table uses the same four verdicts as the method page. The value of naming them is that each one commits the board to something specific. "Continue to monitor" commits it to nothing and is, in practice, a decision to fund.

Three of the four outcomes keep the program running: Proceed as planned, and Pivot or Reset with changes. That matters for how the question is put to management, because asking for an independent read is not asking for a Stop.

Boards often merge the two middle options into one, usually described as "adjusting the plan." They need to stay separate, because they need different amounts of capital and different kinds of oversight.

A Pivot has evidence behind it. The buyer research shows that the problem is real and names the one dimension that fails. Perhaps the price is accepted per line but not per site, or the segment is right but the channel partner does not reach it. The revised plan changes that dimension, and the board can fund it with a clear expectation of what should improve.

A Reset is what the board decides when the evidence cannot answer the question as it was asked. The premise was too broad, or it bundled several claims together, or the program never defined what success would look like. A Reset replaces the premise with a sharper one and funds a test of it, not the full plan.

The risk with a Reset is that it becomes a polite delay. The signs are familiar: the new plan restates the old one with later dates, the success criteria are softer, and the budget is the same. A Reset is only a decision if it produces something testable.

A Reset brief should fit on one page and answer five questions.

What exactly was wrong with the original premise? Not the results. The formulation. For example: "The plan assumed plant managers would buy. Interviews show plant managers want it and cannot approve spending at this level."

What is the new hypothesis? It must be able to fail. "Regional operations directors at mid-size discrete manufacturers will approve a per-site subscription at the planned price within one budget cycle" can fail. "Adoption will accelerate as the product matures" cannot.

What will count as a pass? Stated as numbers before the test starts, and not redefined afterward.

What is the capped budget for the test? Enough to answer the question, and no more. The full rollout budget is not released on a Reset.

When is the next decision, and who makes it? A date on the board calendar and a named owner. Without both, a Reset drifts into a continuation.

Writing a Reset this way has a useful side effect. It forces a conversation about what the program was trying to prove, which is often the conversation that did not happen at the deal.

A Stop is the decision boards find hardest to take on a portfolio company's growth bet, and it is often the one that returns the most capital. The difficulty is less financial than social. The program has a team, a sponsor on the management team, and often a place in the story the deal team told the investment committee.

Three things make a Stop easier to take and easier to defend.

The thresholds existed before the evidence. A Stop based on thresholds written after the results reads as a judgment. A Stop based on thresholds the board and management agreed in advance reads as the plan working as designed. This is the strongest single argument for writing pass/fail thresholds into the VCP at the start.

The record separates the premise from the team. A Stop verdict says that the buyers did not behave as the plan assumed. It says nothing about whether the team executed well, and often the team executed well against a premise that did not hold. Saying so explicitly makes it easier to keep good people and move them to the programs that are working. What a defensible Stop record contains is set out in the Stop verdict section of the pilot purgatory paper.

The capital has somewhere to go. A board that stops one program should decide in the same meeting where the released capital goes. A Stop with no redeployment decision tends to be reopened.

The same evidence is worth more early in the hold than late.

In the first year, a Stop or Pivot releases capital while there is time for the replacement program to show results before exit. In the last two years, the options narrow. The program is either part of the exit story or a write-off inside it.

There is a second reason to test early. A buyer's diligence at exit will ask the same questions this paper asks. If a growth initiative in the exit plan rests on untested buyer claims, the buyer's commercial due diligence (CDD) will find the gap and price it. A claim tested and graded during the hold is an asset in the data room. A claim carried untested to exit is a discount. The buyer-side view of this is covered in What the data room cannot show.

The pressure on holding periods makes this sharper. Bain counts 32,000 unsold companies worth $3.8 trillion across the industry, and almost 40% of all companies are now held for more than five years, up from 29% in 2019 [4]. With holds near seven years and a higher growth requirement [1], a program that runs two extra years on a wrong premise costs both the capital and the time to replace it.

The most practical change a board can make is to release capital for new growth bets in tranches, each tied to a graded evidence gate instead of a date.

A gate names the commercial claim that must be true before the next tranche, the pass/fail threshold, the evidence grade required, and the date the board will review it. A typical sequence might look like this, with the thresholds set for each bet:

1. Before the first tranche: the problem and the buyer. Scored interviews confirm urgency and budget authority in the target segment.
2. Before the second tranche: the price and the channel. Buyers confirm willingness to pay at the planned price, and the route to them works at the planned cost.
3. Before the scale tranche: repeatability. The results from the first accounts or sites hold at the next ones.

This structure fits the 100-day plan and the quarterly board cycle. The 100-day plan names the growth bets and their first gates. Each quarterly meeting reviews the gates that fall due. The claims still open at close from the deal diligence become the first gates, with the same thresholds.

A gate passed on benchmarks or a handful of interviews is a weaker basis for the next tranche than one passed on scored interviews with the budget holders. The board can accept the weaker case, knowingly, with a smaller tranche.

The input the board is missing is buyer evidence gathered from outside the program. In practice that means structured interviews with the buyers the plan depends on: the budget holders, the people who would run the system, and the functions that can stop a purchase. Where the buyer is a plant, that usually means the plant manager or VP of operations, the plant controller who checks the value case, the maintenance and reliability manager, and the EHS and procurement teams. Each interview is scored on the same rubric, and each claim in the plan is graded against thresholds set before the interviews.

Management takes part. They supply context, access to customers and the history of the program, and they see the findings before the board does, so they can correct facts and add context. They do not change the grades, and they do not grade their own premise.

Whether the portfolio company is named to the people interviewed is agreed at scoping. When the interviews are with the company's own customers, naming it usually makes for a fuller conversation. The work is confidential, and the brief, the scoring and the interview record belong to whoever commissioned it, usually the fund or the portfolio company's board.

For AI initiatives specifically, the gap between pilots and paid deployment is wide across the market. S&P Global's 451 Research found that the share of companies that discontinued most of their AI initiatives between proof of concept and production rose from 17% in the fourth quarter of 2023 to 42% in the fourth quarter of 2024 [3]. The same survey estimated that organizations ended 46% of their AI projects on average in the move from a successful proof of concept to production [3]. Gartner forecasts that over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value or inadequate risk controls [5]. None of these figures says anything about a particular portfolio company. They are a reason to test the commercial case of its AI program before the next tranche instead of after it.

On the hypothesis

• What commercial claim about buyers does this program depend on? Can management state it in one sentence that could be proven wrong?
• Was that claim tested with buyers before the deal, or was it a projection?
• Which of the four decisions is management proposing, and what evidence supports it?

On the evidence

• Which inputs to this decision come from inside the program, and which from outside it?
• Have the budget holders in the target segment been interviewed, or only the champions and existing customers?
• For each claim, is the evidence graded Verified, Benchmarked or Assumption?

On a Reset

• What is the new hypothesis, and how could it fail?
• What is the pass criterion, written as a number?
• What is the capped budget, and when is the next decision?

On a Stop

• Were the thresholds for stopping agreed before the evidence came in?
• Does the record separate the premise from the team's execution?
• Where does the released capital go, and who owns that decision?

On timing

• How many years remain in the hold, and how long would a replacement program need to show results?
• Will this program be part of the exit story? If so, which of its claims would a buyer's diligence test, and what would it find today?

1. Bain & Company, "Global Private Equity Report 2026," press release, 23 February 2026. https://www.bain.com/about/media-center/press-releases/2026/private-equity-resurgence-gathers-steam-as-new-era-challenges-firms-to-enhance-value-creationbain--company-global-pe-report/
2. 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 16, no. 1 (1976): 27 to 44. https://doi.org/10.1016/0030-5073(76)90005-2
3. S&P Global Market Intelligence, "Voice of the Enterprise: AI & Machine Learning, Use Cases 2025," online survey of 1,006 respondents in North America and Europe fielded 21 October to 25 November 2024. Figures read in the 451 Research Discovery Report "Best practices for delivering AI at scale," May 2025, https://www.verizon.com/business/resources/T51/reports/ai-at-scale-best-practices.pdf. Publisher summary: https://www.spglobal.com/market-intelligence/en/news-insights/research/ai-experiences-rapid-adoption-but-with-mixed-outcomes-highlights-from-vote-ai-machine-learning. As reported by CIO Dive, 14 March 2025: https://www.ciodive.com/news/AI-project-fail-data-SPGlobal/742590/
4. Bain & Company, "Private Equity Outlook 2026: Gaining Traction," Global Private Equity Report 2026, 22 February 2026. https://www.bain.com/insights/outlook-gaining-traction-global-private-equity-report-2026/
5. Gartner, "Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027," press release, 25 June 2025. https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027

What is not yet sourced

• The S&P Global figures (reference 3) were read in S&P's own May 2025 Discovery Report and match CIO Dive's reporting. The full Voice of the Enterprise study is paywalled and was not reviewed.
• The Gartner figure (reference 5) is a forecast, not a measurement.
• Staw's finding (reference 2) comes from a laboratory simulation of a business investment decision with student participants. Applying it to deal teams and boards is our reading, not his result.
• The four ways a VCP growth bet stalls, the tranche-and-gate sequence, and the observation that an untested growth claim is discounted at exit rest on Thing Company's experience, not a published study.
• The four verdicts and the Verified, Benchmarked and Assumption grades are Thing Company methodology.

Thing Company is an independent market validation practice for industrial technology. We test whether a buyer exists at a price that works. Each engagement is led by a senior practitioner from Thing Company or its expert network.

For a portfolio company, a Sprint gives the board an independent read before the next tranche. For investors sets out how it fits a deal or a hold. Facing a funding vote on a stalled growth bet? Start with a Sprint Assessment.

Not ready to talk? Run the free stall diagnosis self-check from the toolkit.