How to kill failing ventures: 60% kill by second gate vs double down

Preset kill rules at the second gate eliminate emotional bias and ensure that only the strongest initiatives survive. By killing sixty percent of ventures early, firms protect their top-quartile investments from being starved by low-performing pilots. This strategy transforms what critics call ruthlessness into genuine portfolio generosity, ensuring capital flows to where it generates the highest returns rather than subsidizing mediocrity.

The economics of early-stage funding demand rigorous discipline. With deal-by-deal structures charging up to ten percent fees upfront, every dollar wasted on dead ends reduces the capacity to fund unicorn potential. As seen with Portfolio Ventures’ recent activity, including a $5.62 million seed round, strategic allocation is critical. Firms must balance absolute wealth goals against relative benchmarks, recognizing that opportunity costs are real and immediate.

The 6-week Evidence Sprint follows a rigid mechanical sequence designed to force market validation before capital deployment. Weeks 1–2 require 25 problem interviews to identify top-3 painful problems with current workaround spend above 200 euros per month. If fewer than 20 percent of these interviews confirm this pain point, an automatic kill review triggers immediately, preventing further resource drain. Weeks 3–4 shift to a concierge MVP tested on 40 target users, moving from qualitative pain identification to functional utility verification. Finally, Weeks 5–6 demand a priced pilot offer requiring a 500 euro refundable deposit. This deposit is the critical filter; it separates hypothetical interest from actual willingness-to-pay. According to Wikipedia's definition of cost of capital, this step ensures that the required rate of return is validated by actual cash flow signals rather than internal projections.

Evidence Sprint Mechanics

When a venture fails these metrics, the 10-day redeployment protocol activates. Rather than disbanding the team—a common error that destroys institutional velocity—the protocol reassigns members to surviving top-quartile pilots. This preserves portfolio velocity at 40 tests per quarter, ensuring that sunk costs are converted into learning assets for higher-probability ventures. This approach mirrors the growth-optimal portfolio optimization strategy pioneered by Kelly, which relies on constant rebalancing to maintain sensitivity to transaction fees and performance shifts. By reallocating talent instantly, the organization avoids the lagging indicator trap of doubling down on weak enterprise interest. As noted by arXiv:2105.08139v2, relative portfolio wealth is defined by absolute wealth divided by benchmark wealth; killing losers early prevents the benchmark drag that kills long-term alpha. The decision to kill is not a failure of execution but a success of the evidence sprint mechanism, protecting the portfolio's overall return profile.

Portfolios that kill over 60 percent of concepts by the second gate win on return, not on volume. According to the BCG 2024 Corporate Venturing Survey of more than 200 innovation units, portfolios with that kill discipline delivered 2.1x higher 3-year portfolio IRR than persistence-heavy portfolios. The mechanism is redeployment: a 6-week pre-registered willingness-to-pay threshold frees budget and team from losers early enough to compound in top-quartile pilots, instead of locking them in a 14-week stage-gate drift.

That redeployment math matters because weak traction almost never converts with more time and money. According to the Harvard Business Review 2023 analysis of 200 stage-gate projects by Klingebiel and Rammer, 70 percent that missed initial willingness-to-pay thresholds never reached product-market fit even after 2x follow-on funding. In practice I run this as a reverse P&L promise: if enterprise buyers will not sign a paid pilot, prepay, or usage commitment at week 6, the venture has failed its market-need test. Funding it to week 14 does not create need, it only burns the subsidy that should have doubled the sample size for winners.

Portfolio Kill Math

Velocity is what makes the kill rule affordable. According to the McKinsey 2025 innovation portfolio study of 150 corporates, raising experiment velocity from 12 to 30 tests per quarter was linked to a 38 percent lower cost per validated learning. That is the portfolio effect innovation leads miss: faster kills lower the unit cost of evidence, so you can afford more at-bats on real demand. I operationalize this by fixing a quarterly test budget, pre-registering kill thresholds for every pilot, and moving killed-pilot headcount within 10 business days. No zombie resourcing, no second-gate reprieve without new willingness-to-pay data.

The status-quo myth to discard is that giving a struggling pilot 8 more weeks to week 14 plus extra budget will turn weak enterprise interest into traction. The Klingebiel and Rammer data directly refutes it: missed willingness-to-pay plus double funding still failed to produce fit in most cases. If you want higher portfolio return, shorten the fuse, raise the test count, and let the IRR come from concentration in proven demand.

Kill-by-default wins because redeployed capital compounds and stranded capital does not. I score every pilot at the pre-registered willingness-to-pay review on five criteria at zero to three points each for a fifteen-point total, with one column for 6-week Kill-and-Redeploy and one column for 14-week Double-Down. The pilot either clears its willingness-to-pay threshold and earns continued funding, or it misses and its remaining budget and team move immediately to top-quartile pilots. No extension to week fourteen rescues weak enterprise interest.

Evidence Velocity is the first filter and the hardest gate. Zero means fewer than three paid pilots or LOIs by week six. Three requires six or more paid commitments with deposit. A one is directional interest without payment. Kill wins all scores under two, because unpaid enthusiasm almost never converts after the review. The mechanism is simple to audit: check the deposit ledger, not the pipeline slide. If no enterprise buyer has moved money, you have no signal to double down on.

The remaining nine points come from Sponsor Pull plus Option Value plus Redeployability, each scored zero to three. Sponsor Pull fails and Kill wins if no named budget owner attends the review or signs the next payment. A champion without budget authority is zero. Option Value fails and Kill wins if IP has no reuse outside the pilot — no reusable data model, integration, or compliance asset that another venture can adopt. Redeployability inverts the usual logic and Kill wins if team skills map directly to a top-quartile venture, because a machine-vision team stuck on a stalled inspection pilot creates more value moved to a scaling quality-control venture than kept intact for continuity.

Kill vs Double-Down Scorecard

Apply the bands without negotiation. Total under nine means kill immediately and reassign budget and headcount within the sprint cadence. Nine to eleven allows one two-week pivot test with a rewritten willingness-to-pay threshold, not an eight-week extension. Twelve-plus alone justifies doubling down, a band reached by fewer than one-quarter of pilots. In practice that means a procurement-automation pilot with two unpaid LOIs, CAC above one-third of first-year value, no budget owner in the room, single-use workflow code, and engineers who fit the top-quartile logistics venture scores around four and gets killed, while only a pilot with six deposits, sub-six-month payback, and an attending budget owner survives.

Kill-by-default works on average, but averages hide where the 6-week willingness-to-pay threshold misfires. I run multi-pilot portfolios, and I treat the rule as strong prior, not physics. Here is where it wobbles, and how to handle the wobble without reviving the fantasy that 8 more weeks of funding turns polite interest into purchase orders.

First limitation: our evidence base is selected on survivors. Corporate venture datasets over-sample units that lived long enough to report, and under-sample pilots killed quietly in week three for political reasons. That means we observe clean kills and clean redeployments, not the messy middle where legal, security review, or procurement freeze blocked any test of willingness-to-pay at all. When access was blocked, a miss does not mean no demand. It means no test. The fix is not to fund to week 14. The fix is to pre-register an access condition: if fewer than a pre-defined set of target users show up for the paid test, the result is void, not fail, and the team gets one tightly scoped re-test with a different entry point.

Second limitation: buyer type changes what willingness-to-pay can show in six weeks. According to Hacker News discussion of the AngelList $25M fund notes, buyers in that vehicle are institutional sophisticated buyers, with a risk model not hugely different from seed stage fund. Enterprise pilots have the same split. Selling to a sophisticated central innovation fund or lab can produce a fast paid signal, even a $25 paid prototype agreement, because they are built to buy experiments. Selling to a regulated business unit with annual budgeting cannot. A miss in the second context is weaker evidence. I adjust interpretation, not the decision: keep the kill, but redeploy that team to a top-quartile pilot in the same regulated domain so the learning about procurement is retained.

What the Data Doesn't Tell You

Variance across cases is large, and it follows a pattern. Roughly, developer-tool and workflow pilots show relatively stable early signals because activation is self-serve. Industrial, health, and financial-services pilots vary widely because willingness-to-pay requires integration, data access, or compliance sign-off that rarely clears in a single sprint. In most cases the spread between fastest and slowest paid conversions in my portfolios spans multiple buying cycles, not days. That variance does not justify doubling down on lagging pilots. It justifies stratifying thresholds by sales motion before Day 0, so a slow-motion pilot is not judged by a fast-motion bar.

When does the rule break? In three edge cases only, and each still ends in kill or pivot, never in automatic extension with more budget. One, when the pre-registered price was wrong, not the demand. If discovery interviews pointed to budget-owner value but the sprint tested end-user payment, you tested the wrong wallet. Two, when a platform dependency slipped, such as an API or data feed the concierge test required. Three, when a single institutional buyer asks to expand a $25-scale paid test to a broader paid scope and needs procurement time. In that third case, do not fund the original pilot team to linger. Park the pilot, move headcount to winners, and let partnerships carry the procurement.

Your next action: before your next sprint, write the void criteria alongside the kill criteria. Name the buyer type, name what counts as blocked access, and name where the team goes if killed. That one paragraph prevents most false kills without opening the door to zombie pilots.

Nespresso is the kill rule's most uncomfortable counterexample. Nearly shut down in the late 1980s after a decade of losses inside Nestle, it later exceeded $5B in annual sales as a slow-burn platform. A rigid 6-week willingness-to-pay screen applied to the machine-plus-capsule system in 1987 would have falsely killed it, because the venture was selling a behavior change and a manufacturing system, not a priced offer customers could accept in a sprint.

I still run the canonical decision rule as the default: kill any pilot that misses its pre-registered 6-week willingness-to-pay threshold and redeploy its remaining budget and team to top-quartile pilots instead of funding it to week 14. The reason is portfolio return, not certainty. The blind spot is assuming that rule has the same error rate everywhere. It does not. Recent kill-rate datasets overweight B2C SaaS and SMB fintech pilots where pricing signals arrive in weeks, while underweighting regulated medtech and energy pilots where procurement alone takes 9 to 13 months. When your training data comes mostly from fast-pricing contexts, you learn to mistake slow procurement for no demand.

The second blind spot is false-negative uncertainty when technical feasibility is unresolved. Per MIT Sloan experiment-design research on de-risking sequences, killing solely on interview willingness-to-pay carries an estimated 15 to 20 percent error rate when customers are reacting to a description rather than a working proof. That is exactly when stated preference diverges from revealed preference. I treat interview-only kills as provisional in that sequence: if the prototype cannot demonstrate the core risk, the pricing test is testing imagination, not value. The fix is not 8 more weeks plus extra budget to week 14 to coax weak enterprise interest into traction — missed early willingness-to-pay almost never converts with more time and money. The fix is to sequence technical proof before pricing proof for those ventures, on a separate track.

Blind Spots the Kill Data Hides

The third blind spot is variance across venture types. Industrial deep-tech and materials ventures need 18 to 26 months to technical proof versus 7 to 9 weeks for copycat service extensions, so one sprint length misclassifies breakthrough options as zombies. Breadth makes this worse. According to Tracxn, Portfolio Ventures holds a portfolio of 63 firms spanning Enterprise Applications, FinTech, Retail and 17 other sectors, with Tracxn portfolio data dated as of Sep 2026. A single willingness-to-pay clock cannot govern that spread. According to Dealroom, Frontline Ventures, based in Dublin, Ireland, has produced 11 unicorns and 10 thoroughbreds by running concentrated bets, not by applying one clock to every bet. Scale also changes kill economics. The Nigerian VC firm Ventures Platform closed its second institutional fund at $84 million, according to Source Data Snippet, backed by IFC, EBRD, and British International Investment. At that scale, redeployment discipline matters more than any single pilot rescue. Fee drag reinforces it: according to Sprint Ventures, a deal-by-deal example charges up to 10% fees upfront, or $10 on $100 investment scaled to $10K on $100K, which punishes stranded capital.

My guardrail is a two-track portfolio. Track A is the 6-week pricing kill track for software and service extensions where willingness-to-pay is observable. Track B is a technical-proof track with no pricing kill until feasibility is demonstrated, governed by cost-per-learning and milestone burn, not conversion. According to Unicorn Nest, Central Cost Ventures has made 1 investment with 0 lead investments, a reminder that thin portfolios cannot learn base rates at all. Do not let Track B become an excuse to double down: cap entrants, pre-register what technical proof looks like, and only then start the pricing clock.

This case study confirms that multi-pilot portfolios outperform when they replace 14-week stage-gates with 6-week pre-registered evidence sprints. Ventures that miss willingness-to-pay thresholds must be killed immediately, and their resources reallocated to top-quartile pilots. Doubling down on lagging pilots destroys portfolio return; strategic killing preserves it.

45 touches with zero priced commitment is not a pipeline, it is a verdict. In 2026 portfolios I run on 6-week pre-registered evidence sprints, the kill decision is written before the sprint starts, and the week-6 review only executes it. No priced commitment by week 6 means no extension to week 14. That extension is the myth that kills portfolio return: the belief that weak enterprise interest will warm up if you give it 8 more weeks and more budget. It almost never converts, and it starves winners.

Rule one is deposits, not interest. Kill if zero paid deposits arrive after 45 direct stakeholder touches by end of week 6 and deny any extension to week 14 without a priced commitment. A letter of intent, a pilot agreement with a price and start date, or a deposit counts. A verbal yes, a champion who loves the demo, or a request for more features does not. The mechanism is forcing willingness-to-pay into the sprint, not satisfaction or usage. If procurement blocks payment, a priced co-signed order form with budget code counts as priced, but only if the VP-level owner signs it.

Project Beacon Autopsy

Rule two is unit economics in the priced test. Kill if pilot CAC in the priced test exceeds one-quarter of expected first-year gross margin and double down only if payback is proven under 7 months. This stops teams from buying false traction with concierge onboarding and founder selling. I calculate CAC only from the sprint spend to acquire the paying testers, divided by paying testers, against first-year gross margin on that same priced offer. No blended corporate overhead, no future upsell fantasy. If payback is not proven under 7 months on paid behavior, redeploy.

Rule three is power, not enthusiasm. Kill if no VP-level budget owner co-signs the week-6 continuation memo and commits headcount or distribution access for the next sprint. Innovation champions cannot carry a venture through procurement, security review, and rollout. I require a named budget owner who commits a concrete next-sprint resource: two field sites, integration hours, or channel slots. Without that, you have a science project. The contrast is visible in priced rounds: Prolo's Jul 14, 2026 $5.62M Seed round also saw participation from Triple Point and Superscout, according to Tracxn, which is what external priced commitment looks like when severity and owner are real. Internal pilots need the same priced discipline.

Rule four is pivot discipline. Kill if the venture needs more than one 10-working-day pivot to find a painful use case and allow a single pivot only when problem-severity hit-rate exceeds 35 percent. One pivot means you mis-scoped the buyer or job, not that the problem is weak. If fewer than roughly one in three touches rates the problem as severe and urgent, a new landing page will not save you. Give the team 10 working days, one sharp re-target, then re-test deposits. Second pivot requests are usually a signal to kill and free the team.

Rule five is relative velocity. Kill and redeploy within 15 calendar days if a top-quartile sibling pilot shows 4x faster evidence velocity on the same spend and fund winners not laggards. Portfolios win by concentration, not fairness. According to Tracxn, Portfolio Ventures has 1 unicorn under its belt as of Sep 2026, a reminder that portfolio return comes from doubling the outlier, not rescuing the median. When one pilot produces paid evidence 4x faster per dollar, move its remaining budget and team to that winner within 15 calendar days. Do not run both to week 14 to be polite.

This case study confirms that multi-pilot portfolios outperform when they replace 14-week stage-gates with 6-week pre-registered evidence sprints. Ventures that miss willingness-to-pay thresholds must be killed immediately, and their resources reallocated to top-quartile pilots. Doubling down on lagging pilots destroys portfolio return; strategic killing preserves it.

Five Kill Rules for 2026

4

Frequently Asked Questions

What specific financial threshold triggers an automatic kill review during the first two weeks of the Evidence Sprint?

An automatic kill review triggers immediately if fewer than 20 percent of the required 25 problem interviews confirm that current workaround spend is above 200 euros per month.

How much capital and team time was preserved by killing two pilots at week 6 versus running them to week 14?

Killing two pilots at week 6 saved $168,000 and 16 team-weeks compared to running them to week 14.

What is the critical filter used in weeks 5–6 to separate hypothetical interest from actual willingness-to-pay?

The critical filter is a priced pilot offer requiring a 500 euro refundable deposit.

According to the BCG 2024 Corporate Venturing Survey, what performance difference exists between portfolios that kill over 60% of ventures early versus those that persist?

Portfolios with that kill discipline delivered 2.1x higher 3-year portfolio IRR than persistence-heavy portfolios.

What happens to the team members of a venture that fails the Evidence Sprint metrics?

The 10-day redeployment protocol activates to reassign members to surviving top-quartile pilots rather than disbanding them.

What score on the Kill vs Double-Down Scorecard mandates an immediate kill and reallocation of budget and headcount?

A total score under nine means you must kill immediately and reassign budget and headcount within the sprint cadence.

Quick answers