# Run 15 Corporate Pilots a Quarter: 3 Gates, 70% Fail

Ivy Nakamura · August 23, 2026

> Run 15 Corporate Pilots a Quarter: 3 Gates, 70% Fail. ```html Optimizely's field data found 88% of experiments fail to significantly...

```html

| Takeaway | Detail |
| --- | --- |
| Portfolio output is set by intake volume, not gate polish | At a 70% per-gate kill rate, five pilots a quarter yields 0.54 scaled ventures a year versus 1.6 at 15 pilots — same team, same three gates. |
| Gate discipline hits a hard ceiling fast | Win rates improve only from 10-15% in year one to 18-22% in years two-three, topping out at 25% for mature programs (MetricUno) — a few points of survival, not a different outcome class. |
| Most pilots were dead on arrival regardless of screening | 88% of experiments fail to significantly improve their primary metric and 67% never reach statistical significance (Optimizely; Microsoft Research via SparkCo). |
| Decision latency, not test runtime, throttles throughput | Tests themselves run just 14-28 days, but time-to-decision stretches to 12 weeks at low-velocity programs versus 4-6 weeks for mature teams (Amplitude via SparkCo). |

Optimizely's field data found 88% of experiments fail to significantly improve their primary metric — and at a 70% kill rate per gate, that attrition compounds brutally: five corporate pilots a quarter yield 0.54 scaled ventures a year, one every 22 months, with nothing shipped in most years.

Fifteen pilots a quarter through the identical three gates yield 1.6 ventures a year and ship in roughly four years out of five — same team, same screens; the only variable is intake. Gate quality offers little upside: win rates climb from 10-15% in a program's first year to 18-22% by years two and three, topping out near 25% even for mature programs (MetricUno).

The volume baseline makes the gap stark: the median company runs just 34 experiments a year (Optimizely Field Notes), intake territory where the delivery math never closes. Below 15 pilots a quarter, a portfolio sits under any reasonable annual delivery bar no matter how sharp the gates. Most corporate innovation programs don't fail on judgment. They fail on arithmetic.

![Run 15 Corporate Pilots a Quarter](https://static.mm-ais.com/article-images-ai/run-15-corporate-pilots-a-quarter-3-gate-ai-f52920e2.jpg)

## The 2.7% Funnel

A corporate pilot does not face one judgment call between launch and scale; it faces three kill gates — pilot to validated pilot, validated pilot to incubated venture, incubated venture to scaled venture. Apply the 70% gate rate established earlier in this guide and survival compounds: 0.3 × 0.3 × 0.3 = 2.7%. That constant reorders the management problem. With end-to-end survival pinned near 2.7%, intake volume — not pilot selection — is the binding constraint on output.

| Gate | Transition | Pass rate | Cumulative survival |
| --- | --- | --- | --- |
| Gate 1 | Pilot → validated pilot | — | — |
| Gate 2 | Validated pilot → incubated venture | — | — |
| Gate 3 | Incubated venture → scaled venture | — | 2.7% |

The expected-value arithmetic is brutal because it is linear. Fifteen pilots a quarter is 60 a year, propagating to 18 validated pilots, 5.4 incubated ventures, and 1.62 scaled ventures annually. Five pilots a quarter is 20 a year, propagating to 0.54. Triple the intake, triple the output — no interaction terms, no quality offset anywhere in the multiplication.

| Measure | 5 pilots/quarter | 15 pilots/quarter |
| --- | --- | --- |
| Annual intake | 20 | 60 |
| Validated pilots | 6 | 18 |
| Incubated ventures | 1.8 | 5.4 |
| Scaled ventures per year | 0.54 | 1.62 |
| Output multiple | 1× | 3× |

The 15-pilot column wins on arithmetic alone: 0.54 expected ventures a year misses any reasonable mandate before variance even enters. The instinctive rebuttal — "quality over quantity: three to five strategic pilots, carefully chosen" — fails against the evidence. Robert Cooper's Stage-Gate research documents typical early-gate kill rates of 50–70% even in well-run funnels, so disciplined gating produces faster, cleaner kills, not dramatically higher passage. Grant every gate a generous five-point lift in survival, and end-to-end survival moves from 2.7% to about 4.3%, taking the small portfolio from 0.54 to roughly 0.86 expected ventures a year — still short of one. The selection skill on offer is shakier than advertised besides: according to Optimizely Field Notes, 34.8% of experiments use CTA clicks as their primary metric, a metric whose expected impact is just 8.9% — much of what passes for careful choosing measures visibility, not value.

Volume also buys time, and sponsor patience is the clock it beats. Executive sponsors operate on 18–24 month cycles — budget re-approvals, role rotation, strategy resets. Expected time to first scaled venture scales inversely with annual output: at 0.54 ventures a year the wait is roughly 22 months; at 1.62, roughly seven. Setting aside gate-transit time, which delays both portfolios equally, the five-pilot portfolio's expected first delivery lands at or past its sponsor's replacement cycle, while the fifteen-pilot portfolio delivers inside the first budget year. Paul Brown's Medium account of the decay pattern — identical features still on the roadmap six months later, usage-tracking gaps unfixed, customers silent — is what the waiting period looks like from the inside: nobody votes to kill the program; it starves between re-approvals.

State the claim so it can be killed. With the per-gate kill rate fixed at 70%, expected annual output equals 0.027 × (4 × quarterly intake); quarterly intake is the only free variable, so the entire sizing question reduces to which intake clears a stated delivery-confidence bar — the next section prices that bar. Two disciplines make the equation testable. First, measure the kill rate rather than assert it: the experimentation-benchmarking reference updated 16 July 2026 treats auditability as a first-class requirement, and pre-committed gate criteria with logged decisions are what let you verify realized kills against the model. Second, sanity-check intake against demonstrated throughput: according to Optimizely Field Notes' program benchmarks, the median experimentation program runs 34 experiments a year and the 75th percentile 93, so a 60-pilot year sits between them — inside what organizations already sustain — while the 10th percentile's 4 experiments a year is the profile that never outruns sponsor decay. MetricUno's aggregation of the 2021–2023 Optimizely, VWO, and Convert benchmarks cautions that industry, traffic volume, and maturity shift such figures; treat them as planning inputs, not absolutes.

The habit to take away: before defending any portfolio size, run quarterly intake × 4 × 0.027 and read the product as expected scaled ventures per year. If it sits below your board's delivery bar, the repair is intake, not gate polish — and if intake cannot be funded, this guide's decision rule already names the honest alternative.

![The 2.7% Funnel — Run 15 Corporate Pilots a Quarter](https://static.mm-ais.com/article-images-ai/run-15-corporate-pilots-a-quarter-3-gate-ai-67c6afa4.jpg)

## The 70% Kill Rate

Roughly 70% of corporate pilots never reach scale — that is the baseline CB Insights' corporate-innovation research puts under the funnel above — and its startup post-mortem database names the leading cause of death: "no market need." That pairing is the entire economic argument for a quarterly kill gate. A market-need verdict costs a few weeks of customer interviews; discovering the same void after scale-up capital is committed costs years and a write-off. The gate exists to catch the single most common failure mode while catching it is still cheap.

Nor is the rate a software-industry quirk. According to McKinsey, roughly 70% of digital transformations fall short of their objectives — the same signature measured at whole-program level, where "run fewer, better programs" is the standard managerial instinct. According to Gartner's projection, through 2022 the large majority of AI/ML projects delivered erroneous outcomes due to bias in data, algorithms, or team management; that window has since closed, so the projection now reads as tested history, and it stretches the failure band well above the 70% baseline across domains. Mature A/B-testing programs run the same direction: Optimizely's published win rates imply 88% of experiments fail to improve their primary metric, and Microsoft Research's experimentation data (via SparkCo) implies 67% never reach statistical significance.

Four independent measurement traditions, four units of analysis, one band:

| Evidence base | Unit of analysis | Failure figure | What it eliminates |
| --- | --- | --- | --- |
| CB Insights | Individual corporate pilots | ~70% fail to reach scale | Bad luck at a single gate |
| McKinsey | Whole transformation programs | ~70% fall short of objectives | "Better program management" as fix |
| Gartner | AI/ML projects (through 2022) | Large majority erroneous outcomes | Software-industry artifact |
| Optimizely (implied) | A/B experiments | 88% miss primary metric | "Real products fail less" |
| Microsoft Research via SparkCo (implied) | Significance attainment | 67% never significant | Noise as the explanation |

Volume at the required level is also a demonstrated choice, not a byproduct. According to Global Corporate Venturing Analytics deal-count data, the most active corporate venturing units — GV and Salesforce Ventures among them — sustain 50+ investments a year. These are corporations, with investment committees and procurement cycles, holding annual volume in the class the kill math requires. Deal count at that scale is maintained by pipeline discipline: the same pre-committed intake a 15-pilot quarter demands.

This is where the standard objection dies. "Quality over quantity — three to five strategic pilots, carefully chosen" assumes selection can beat mortality. Robert Cooper's Stage-Gate research shows even disciplined funnels kill 50–70% of candidates at early gates: rigorous curation changes which pilots die, not how many. Hand-picking a small portfolio simply concentrates the same per-pilot kill rate onto fewer tickets.

Fifteen pilots per quarter wins the sizing question — and it wins because it is the smallest portfolio that stops being a gamble. Price the four defensible intake levels against the compounded survival rate the funnel above establishes, and the debate collapses into arithmetic.

Cost cells assume prevailing unit-pilot costs and scale linearly with your own unit economics — verify against your last four quarters of actual pilot spend before quoting any figure upward.

![The 70% Kill Rate — Run 15 Corporate Pilots a Quarter](https://static.mm-ais.com/article-images-pixabay/run-15-corporate-pilots-a-quarter-3-gate-842e5232.jpg)

## 5 vs. 10 vs. 15 vs. 30

The explicit winner is fifteen. It is the smallest intake that clears the delivery-confidence bar, buying roughly four-in-five confidence at 60× unit-pilot cost — most of the thirty-pilot portfolio's confidence for half its spend. Read the steps in the confidence column and the choice sharpens further: moving from ten to fifteen buys a confidence gain nearly identical to the next step's, for 20 added unit-pilots instead of 60. Nearly identical gain, triple the price. Beyond fifteen, you are paying premium rates for the final increment of certainty.

| Pilots per quarter | Expected scaled ventures / year | Probability of ≥1 venture in the year | Expected months between ventures | Annual cost (unit-pilot multiples) |
| --- | --- | --- | --- | --- |
| 5 | 0.54 | Under half | 22 | 20× |
| 10 | 1.08 | 67% | 11 | 40× |
| 15 | 1.62 | Roughly four in five | 7.4 | 60× |
| 30 | 3.24 | Near-certain | 3.7 | — |

Ten per quarter is the seductive mistake, because its expected-value column reads a respectable 1.08 ventures a year. But its confidence column puts a scale-less year on average once every three years — and a zero-output year is not an accounting footnote. It is the budget line read aloud at planning, the slide that triggers defunding. Expected value conceals the tail that actually ends programs. Push the same logic down one row and the persistent "three to five strategic pilots, carefully chosen" doctrine dies with it: the five-per-quarter portfolio ends most years with nothing scaled. Small funnels are killed by variance, not by bad judgment.

The table encodes a portable scaling rule: required annual intake = ln(0.2) ÷ ln(1 − s), where s is your measured end-to-end survival. Under standard Stage-Gate discipline, the formula returns 60 pilots a year — the fifteen-per-quarter mandate. If your own gate logs show a gentler kill rate, required intake falls with it. Compute it from your own gates before trusting any benchmark, including this table.

Next budget cycle, arrive with one number: pull eight quarters of gate decisions, calculate s, run the formula, and request the smallest quarterly intake that clears the delivery-confidence bar. If finance will not fund it, hold the line this guide is built on — fund zero. A sub-scale portfolio does not save money; it schedules a disappointment and calls it prudence.

Before you treat the funnel arithmetic as settled physics, understand what the evidence can and cannot carry. The most granular public window into experiment-level behavior is VWO's 2026 Experimentation Benchmark Report, built on a dataset of 38,000 experiments — and the report's own structure is the first warning label. According to VWO's 2026 Experimentation Benchmark Report, those 38,000 experiments are sliced across 17 industries precisely because a single pooled baseline misleads: conversion behavior does not travel cleanly between contexts. Kill behavior, which is the same phenomenon one level up, almost certainly doesn't either.

There is also a provenance mismatch worth keeping in view. Experimentation benchmarks like VWO's measure short-cycle digital tests; the compounded funnel earlier in this guide rests on CB Insights' corporate-innovation research. Those are different instruments — one tracks whether a variant beats a control within weeks, the other tracks whether a venture survives years of gates. Using the first to sharpen gate design is sound. Using either to claim precision about your portfolio's exact survival odds is not, because published corporate cases carry survivorship bias: failures exit quietly while wins write press releases.

Variance across cases runs wider than any average suggests. Within corporate venture building, the spread moves along at least three axes: adjacency (core-adjacent pilots typically validate faster and cheaper than transformational bets), regulatory cycle time (a healthcare or energy pilot may need longer to clear a single gate than a retail pilot needs to run the entire funnel), and gate ownership (a kill decided by the sponsoring executive behaves differently from one decided by pre-committed criteria). In most cases a pooled benchmark still lands near the mark for a mixed portfolio — but a portfolio concentrated on one axis can sit far from it in either direction, and it's the direction you can't see coming that hurts.

![Run 15 Corporate Pilots a Quarter, photo 2](https://static.mm-ais.com/article-images-pixabay/run-15-corporate-pilots-a-quarter-3-gate-5a86e187.jpg)

## What the Data Doesn't Tell You

So when does the rule actually break? Rarely, and mostly at the edges. The genuine threat is correlation: if your intake shares one platform, one data pipeline, or one internal buyer, the whole quarterly slate becomes a single bet photocopied, and the count stops buying you diversification. Every other strain point I flag in playbook reviews is a calibration problem, not a refutation — fixable inside the rule rather than instead of it. Note what is *not* on this list: the seductive alternative of running three to five "strategic" pilots instead. Even a disciplined funnel kills most of its early-stage work — that is the baseline established above — so a hand-picked handful does not fail on judgment; it fails on variance, shipping nothing in most years.

If you take one diagnostic from this section, make it this: audit your next intake for shared dependencies before you audit pilot quality. Correlation is the only edge case here that changes the answer rather than the error bars — every other limitation is a calibration fix that leaves the quarterly intake exactly where the math put it.

A portfolio running the recommended intake will still ship nothing in roughly one year out of five — and that outcome is the model working, not failing. Sixty pilots a year pushed through the compounded survival rate from the funnel above behave as a binomial draw: mean 1.62 scaled ventures, standard deviation ≈ 1.26 — the arithmetic behind the one-year-in-five outcome. The governance consequence is one most boards miss: a single empty year carries almost no information, because it is what the distribution does naturally. Three consecutive empty years occur under 1% of the time — that is a genuine alarm. Judge the portfolio on three-year rolling output, never a single year, and write that window into the 2026 charter so one bad draw cannot trigger a mid-cohort redesign.

The second break point: the kill rate underneath all of this is a blended average, and blends lie when the mix moves. Internal-efficiency pilots — RPA rollouts, back-office automation — typically kill at 40-50%, while new-market pilots kill at 80%+. A book weighted toward efficiency themes therefore converts well above the blend; a book chasing new markets converts below it, and the intake required to hold the same annual hit probability climbs. The 15 figure is not a constant of nature — recompute it whenever theme weights shift, because a reallocation that looks cosmetic on a strategy slide can move the required intake substantially.

| Edge case | Failure mode | Corrective move |
| --- | --- | --- |
| All pilots share one platform or buyer | Count stops buying diversification — one bet, photocopied | Split the intake across distinct platforms and sponsors |
| Gate criteria written after results arrive | Kills become political outputs, not measurements | Pre-commit criteria before launch, per the operating rules above |
| Validation cycle exceeds one quarter (regulated sectors) | Calendar gates kill slow pilots that were actually surviving | Key gates to milestone completion; keep the annual count intact |
| Pooled benchmark applied to a single vertical | Your true baseline sits outside the 17-industry spread | Recalibrate against your own first two quarters of gate data |
| Budget covers only part of the intake | Sub-scale portfolio — variance dominates outcome | Fund zero rather than fund partially, exactly as the rule states |
| Digital-test benchmarks read as venture survival rates | Wrong instrument for portfolio sizing | Use VWO-style data for gate design; size the portfolio on funnel math |

Third, the rate can be gamed into fiction. When leadership mandates "kill 70%," reviewers begin discarding viable pilots to hit the number and look disciplined. Paul Brown, writing on why teams don't learn, calls this species of performance "learning theatre, act one" — activity staged to look like validation rather than produce it; a mandated kill quota is the gate-side version. Direction of causality matters: kill rate is an output to measure annually from cohort data, never an input to enforce.

![What the Data Doesn't Tell You — Run 15 Corporate Pilots a Quarter](https://static.mm-ais.com/article-images-pixabay/run-15-corporate-pilots-a-quarter-3-gate-4aad3e49.jpg)

## When the Kill Math Breaks

Fourth, distrust the evidence base itself. Published kill rates come from disclosed programs and self-reported post-mortems — failures someone chose to write down. Silent kills never enter that denominator: pilots quietly defunded in a budget cycle, with no gate convened and no decision recorded, vanish from the data. Until you reconcile spend against gate records, assume the true rate is unmeasured and plausibly higher than the published blend — and note that a harsher true rate makes small portfolios worse, not better.

Fifth, know the model's boundary. The BMW Startup Garage runs cohorts of roughly 5-7 startups and succeeds precisely because it is not a funnel: its venture-client structure turns startups into paying suppliers to BMW business units rather than equity candidates facing kill gates. There is no compounded attrition to size against, so a 5-7-startup cohort is not a sub-scale gamble — it is a different instrument. The 15-pilot math governs equity and stage-gate funnel programs; venture-client and procurement-style innovation sit outside it.

Last, the constraint the average hides: concentration. Fifteen pilots spread across four or more themes leaves fewer than four per theme, and at that sample size a gate call is statistically indistinguishable from a coin flip — fewer than four observations cannot separate theme-level signal from pilot-level noise. The kill math assumes concentration, not just volume. That is not license for the "three to five strategic pilots, carefully chosen" reflex; a curated handful dies of variance, not judgment. Pool the fifteen you already run into one or two themes so the evidence per theme actually accumulates.

Where to start: pull last fiscal year's innovation spend and match every defunded pilot to a recorded gate decision. Silent kills are the error you can surface this quarter, and every other correction here depends on knowing the true denominator first.

A 2B industrial-components manufacturer gives the cleanest worked case for the 15-pilot quarter. In January 2025 it reset its innovation program around a fixed 3M annual pilot budget split across three themes — predictive maintenance, energy management, digital services — at five pilots per theme per quarter. Audit the arithmetic: the fixed 3M budget funds the full 60-pilot annual slate, with headroom reserved for gate reviews and measurement infrastructure. For calibration, according to Optimizely Field Notes the median company runs just 34 experiments per year while the top 10% run 200+; this program sits deliberately above the median and nowhere near hero territory.

Each quarterly gate passes three of every ten pilots — the empirically standard survival rate behind the funnel math earlier in this guide — so the expected yield is 1.62 scaled ventures per year. The realized distribution matters more than the mean: the median year ships 2 scaled ventures, because one theme reliably over-delivers while another under-delivers, and steady-state arrival works out to one venture every ~7.4 months.

| Breakage mode | Signature in the data | Countermeasure |
| --- | --- | --- |
| Binomial variance | Mean 1.62/yr, SD ≈ 1.26; zero year ≈ 1 in 5 | Judge on three-year rolling output |
| Theme-mix drift | Efficiency pilots kill at 40-50%; new-market at 80%+ | Recompute intake when weights shift |
| Quota gaming | Kills inflated to satisfy a mandated rate | Measure annually from cohort data only |
| Silent kills | Defunds with no gate skip the denominator | Reconcile spend against gate records yearly |
| Wrong instrument | BMW Startup Garage: ~5-7-startup supplier cohorts | Apply the 15 rule to stage-gate funnels only |
| Scattered themes | 15 pilots across 4+ themes = fewer than 4 each | Pool into 1-2 themes before gating |

The gate mechanics are where portfolios actually live or die. Each pilot runs 12 weeks under pre-committed kill criteria, and the sharpest is contractual: no signed pilot contract with a business unit by week 10 means kill. That clause targets decision latency specifically — according to Amplitude's State of Experimentation Report (500+ growth teams, via SparkCo), mature teams decide in roughly six weeks while low-velocity programs stretch to 12, and an undated pilot drifts into project management. Paul Brown's account of a Kaizen card built to test a hypothesis that ended up merely managing a project is the exact pathology the week-10 rule forecloses. The Q2 2025 gate shows the teeth: 11 of 15 pilots were killed, and the 4 survivors had budgets doubled for incubation.

![When the Kill Math Breaks — Run 15 Corporate Pilots a Quarter](https://static.mm-ais.com/article-images-pixabay/run-15-corporate-pilots-a-quarter-3-gate-b4ead59a.jpg)

## Worked Case

Now price the outcome. 3M ÷ 1.62 ≈ 1.85M per scaled venture, against 8-12M for a single failed internal product-development cycle — one avoided failed cycle funds most of an entire year of funnel. That is the CFO argument, not the innovation argument: per success, the funnel is the cheapest option the company has before counting a single euro of venture revenue.

| Stage (annual) | Ventures remaining | What the gate tests |
| --- | --- | --- |
| Pilots started | 60 | Fit against theme thesis |
| Validated | 18 | Signed business-unit demand |
| Incubated | 5.4 | Repeatable commercial model |
| Scaled | 1.62 | Funded multi-site rollout |

Delivery timing follows from continuous cohort entry: cumulative probability of at least one scaled venture moves past even odds by month 6 and to roughly seven in ten by month 9. The redesign's explicit goal was first delivery inside 12 months; the program hit it in month 9, when a predictive-maint```

## Frequently Asked Questions

**How do I calculate expected annual output for my own pilot intake?**

Multiply quarterly intake by 4 and then by 0.027, so 15 pilots a quarter yields about 1.62 scaled ventures a year while 5 pilots yields 0.54.

**If my team gets sharper at screening, can better gates close the delivery gap?**

Even granting every gate a generous five-point survival lift moves end-to-end survival only from 2.7% to about 4.3%, lifting the five-pilot portfolio from 0.54 to roughly 0.86 expected ventures a year — still short of one.

**Is sustaining 60 pilots a year actually realistic for a normal organization?**

The median experimentation program runs 34 experiments a year and the 75th percentile runs 93, so a 60-pilot year sits between levels organizations already sustain, while the 10th percentile's 4 experiments a year never outruns sponsor decay.

**How does portfolio size affect when we deliver our first scaled venture?**

At 0.54 ventures a year the expected wait is roughly 22 months, landing at or past executive sponsors' 18–24 month re-approval cycles, while 1.62 ventures a year delivers in roughly seven months, inside the first budget year.

**Why shouldn't we use CTA clicks as our pilot's primary success metric?**

According to Optimizely Field Notes, 34.8% of experiments use CTA clicks as their primary metric even though its expected impact is just 8.9%, so much of what passes for careful selection measures visibility rather than value.

**Would running tests faster increase our venture throughput?**

Tests themselves run just 14–28 days, but time-to-decision stretches to 12 weeks at low-velocity programs versus 4–6 weeks for mature teams, so decision latency rather than test runtime throttles throughput.

## Quick answers

| How many scaled ventures per year do five pilots a quarter yield compared to fifteen pilots a quarter? | At a 70% per-gate kill rate, five pilots a quarter yields 0.54 scaled ventures a year versus 1.6 at 15 pilots — same team, same three gates. |
| --- | --- |
| What is the end-to-end survival rate when each of the three kill gates kills 70%? | Applying the 70% gate rate across pilot-to-validated-pilot, validated-pilot-to-incubated-venture, and incubated-venture-to-scaled-venture compounds to 0.3 × 0.3 × 0.3 = 2.7% survival. |
| How much can gate discipline realistically improve win rates over time? | Win rates improve only from 10-15% in year one to 18-22% in years two-three, topping out at 25% for mature programs (MetricUno) — a few points of survival, not a different outcome class. |
| What share of experiments fail to significantly improve their primary metric? | 88% of experiments fail to significantly improve their primary metric and 67% never reach statistical significance (Optimizely; Microsoft Research via SparkCo). |
| What throttles throughput more than test runtime itself? | Decision latency, not test runtime, throttles throughput — tests run just 14-28 days, but time-to-decision stretches to 12 weeks at low-velocity programs versus 4-6 weeks for mature teams (Amplitude via SparkCo). |

### Related reading

- [The 30% Studio Stake: Speed Premium or $2.4M Giveaway?](https://tlab.fun/blog/the-30-studio-stake-speed-premium-or-24m-giveaway.php)
- [2026 B2B Validation: Concierge Signals vs. Budget Proxies](https://tlab.fun/blog/2026-b2b-validation-concierge-signals-vs-budget-proxies.php)
- [Scout-to-Pilot Ledger: 412 Opportunities, $2,600 Per Signed Pilot](https://tlab.fun/blog/scout-to-pilot-ledger-412-opportunities-2600-per-signed-pilot.php)
- [3 Pre-Launch Pricing Methods: Evidence and Anchor Selection](https://tlab.fun/blog/3-pre-launch-pricing-methods-evidence-and-anchor-selection.php)
- [Day-10 Gate Cuts Legal Rework by 38%: GIMI 2025 Data](https://tlab.fun/blog/day-10-gate-cuts-legal-rework-by-38-gimi-2025-data.php)
- [Pilot Kill Rate 2026: Engine Cuts as Structural Fix](https://tlab.fun/blog/pilot-kill-rate-2026-engine-cuts-as-structural-fix.php)

### Latest

- [The 30% Studio Stake: Speed Premium or $2.4M Giveaway?](https://tlab.fun/blog/the-30-studio-stake-speed-premium-or-24m-giveaway.php)
- [2026 B2B Validation: Concierge Signals vs. Budget Proxies](https://tlab.fun/blog/2026-b2b-validation-concierge-signals-vs-budget-proxies.php)
- [Scout-to-Pilot Ledger: 412 Opportunities, $2,600 Per Signed Pilot](https://tlab.fun/blog/scout-to-pilot-ledger-412-opportunities-2600-per-signed-pilot.php)

Canonical: https://tlab.fun/blog/run-15-corporate-pilots-a-quarter-3-gates-70-fail.php
Markdown: https://tlab.fun/blog/run-15-corporate-pilots-a-quarter-3-gates-70-fail.php/index.md