Original Research · Arc Shift Ventures

The build isn’t the hard part anymore.
The first external dollar is.

Corporates and consultancies have industrialised venture building. Sprints are faster, archetypes are mapped, GenAI has collapsed prototype cost. And yet the ventures coming off these lines keep stalling in the same place: first real revenue — the first dollar paid by someone who isn’t the parent. We traced every published venture-building success statistic we could find back toward its primary source and sorted them by what they actually measure. The optimistic numbers all measure something upstream of revenue. The revenue numbers are the bleak ones.

Talk about a venture you’re building See the numbers

Four numbers that frame the problem

22%

of new businesses launched by companies in the past decade have successfully scaled. The rest launched — and then didn’t.

A global consultancy’s annual venture-building survey (disclosed method), 2021 · Grade A · survey-based, self-reported

4 in 5

new corporate-built businesses still haven’t reached US$50M annual revenue four or more years after launch.

Same global consultancy survey, 2021 · Grade A · survey-based, self-reported

154

venture studios have ceased operations out of ~1,107 ever established — and on Q3 data (17 new registrations against 20 closures), 2024 was on track to be the first net-decline year.

Big Venture Studio Research 2024 (independent census) · Grade B · cumulative to Sept 2024, not an annual figure; the net decline was the report’s projection, not a confirmed year-end result; closure = announcement or prolonged site inactivity

published venture-studio or venture-build success statistics, out of every one we audited, that measure third-party revenue. They measure launches, funding rounds, or self-assessed “success” — never external customers paying.

Original finding · Arc Shift audit, 2026 · Grade EST · bounded, falsifiable claim

The core finding

Sort the industry’s numbers by what they measure, and the contradiction disappears

Read the venture-building literature cold and it seems to disagree with itself. Studios say the model works brilliantly; the survey data says most corporate ventures never scale. Both are reporting honestly. They’re just measuring different rungs of the same ladder — and the numbers get worse the closer the metric sits to a third-party customer paying real money:

“Venture launched / concept validated” — one major strategy consultancy reports its corporate venturing model “achieves success 66% of the time” (vendor-measured, own portfolio — Grade B). Another cites 70+ ventures built (a count, not a rate — Grade B).

“Funding raised” — a global venture-studio association’s member survey reports 84% of studio startups raise seed, and 72% of those reach Series A (member self-reports — Grade B; see citation audit).

“Meets expectations for scale and growth” — the field’s largest annual global survey: roughly 44–50% of respondents, and only executives at “expert” serial builders hit high rates (self-assessed — Grade A for method, soft for meaning).

“Actual revenue” — only 22% scaled; four in five below US$50M after four or more years (Grade A).

“The building vehicle itself survives” — 154 cumulative studio closures as of Sept 2024; several blue-chip corporate venture units shut in the same period (Grade B).

The falsifiable claim at the centre of this page: no published venture-studio or corporate-venture-build success statistic measures external revenue with a disclosed sample. If you can find one, I will amend this page and credit you at the top of it.

Why does the ladder thin out exactly there? Because in a corporate build, the first customer is usually the parent — and revenue from your parent is a transfer, not a test. It arrives without a procurement gauntlet, without a competitive alternative, without churn risk. It proves the sponsor’s authority, not the market’s demand. One major consultancy’s own April 2025 archetype framework lists “a ready ‘first customer’ within its own corporation” as an endowment — which it is. It is also the single most reliable way to reach month eighteen with revenue on the books and no evidence anyone else will ever buy.

Where it actually breaks

Not in the sprint. In the eighteen months after it.

The funded scope ends before the market test begins. Look at the best-designed programme in the region — a Singapore government venture-build co-funding scheme with S$32M committed and nine appointed partner firms, global strategy consultancies and specialist venture studios among them. Co-funding covers the six-month concept-validation sprint through to an “investible business plan”; on the startup-partnership track, support runs “up till the pre-pilot stage.” That is exactly what a government programme should fund. But it means every actor in the chain — studio, sponsor, programme — is paid and measured on milestones that all sit before first external revenue. Nobody’s economics break if the venture launches and stalls. Except the venture’s.

The parent’s order is priced by politics, not procurement. When the parent is the first customer, the deal is typically struck sponsor-to-sponsor at a number that makes the venture’s early P&L presentable. No external buyer will ever pay that way. The venture learns to sell to an organisation that cannot say no — then meets its first one that can.

Internal revenue masks the PMF gap that the build phase skipped. The global survey data is blunt: novice builders are three times more likely than experts to build a venture without product-market fit. A parent-as-customer launch doesn’t correct that error — it defers its discovery by a year or more, at full burn. GenAI has made this worse, not better: the consultancies’ own 2025 argument is that build cycles have collapsed. Building got cheap. Validating became the expensive thing to skip.

The venture’s Year-1 revenue isn’t in anyone’s operating plan. This is the pilot-purgatory mechanism from our pilot research, transposed. The parent business unit that is supposed to become the anchor customer locked its budget six to twelve months before the venture launched. The “committed” first order has to displace a line item someone else already owns.

The vehicle can die before the venture does. 2024 was the first net-decline year for studios globally, and corporate venture units closed in a steady drumbeat through 2023–25 — most often, per trade-press reviews, because the parent restructured, not because the portfolio failed. A venture whose only customer and only funder are the same distressed parent has concentrated every risk in one balance sheet.

What actually works

The saves are structural — and set up before the build, not after the stall

The pattern across the credible evidence — the global survey’s expert-builder factors (Grade A), the structural design of the better studios, and three corporate spin-outs I’ve operated through this exact transition — is the same one we found in pilot conversion: the saves are structural, and they’re set up before the build, not after the stall.

If you’re the corporate

Make the parent buy like a stranger. If the parent is the first customer, the order goes through procurement, at a market price an external buyer would recognise, against a written acceptance test. Anything softer is a subsidy wearing a revenue costume.
Cap parent revenue by charter. Write it into the venture’s funding gates: parent revenue falls below 50% by month 18 (pick your number — but pick one before launch). A captive-revenue cap is the venture-build equivalent of a pilot stop-criterion.
Gate the build on one external proof, not just internal conviction. No seed tranche without at least one external customer through willingness-to-pay validation to a signed order or a dated purchase commitment. The second customer is the real first customer.
Ring-fence the money and the decisions. The survey data: ring-fenced investment raised new-venture odds of success by 28 percentage points, and expert builders are twice as likely to give ventures real decision-making independence. The parent’s job is endowments, not steering.
Name the commercial owner, not just the venture lead. Someone with a name and a diary date owns time-to-first-external-dollar. In our pilot corpus the single strongest intervention was naming the scale owner before day one; it transfers intact.
Measure the programme on external revenue velocity. Launches, sprints completed, and funding raised are inputs. Months-to-first-external-dollar and external-revenue share at month 18 are the outputs. What the innovation team reports upward is what the system will optimise.

If you’re the studio or venture builder

Put first-external-revenue in the mandate — and price it. Tie a fee tranche or an equity vesting trigger to the venture’s first external contracted dollar. At least one studio’s equity-only model already points this direction: a stalled venture is unpaid work. Make that true for every builder on the file.
Refuse “the parent will buy it” as validation evidence. It’s a distribution asset, not a demand signal. Require external willingness-to-pay evidence during the sprint, before build capital burns.
Staff the sell, not just the build. Studio benches are product- and design-heavy; commercial roles are among the most-reported talent gaps in the global survey, for experts and novices alike. A venture architect who has never carried a quota is designing something they’ve never operated.
Sell the sprint with a month 7–18 plan attached. Programme funding ends at the business plan or the pre-pilot stage. The differentiated pitch is what happens after the co-funding stops — channel choice, first-customer acquisition, the parent-to-market revenue transition. That’s the part of the journey the archetype frameworks don’t cover.
Publish a revenue-based statistic. The first studio to publish time-to-first-external-revenue across its portfolio, with a disclosed sample, owns the category’s credibility overnight — precisely because, as of this audit, nobody has.

The citation audit

Four of this field’s most-quoted numbers don’t survive a source check

“84% of studio startups raise seed, and 72% reach Series A”

Attributed to a global venture-studio association

Degraded in transit

The original is the association’s 2020 white paper — a survey of the network’s own member studios reporting on their own portfolios, with no independent verification and a sample described inconsistently across secondary sources (commonly “258 studio startups”). Two problems compound. First, the 72% is conditional — 72% of those that raised seed reach Series A (about 60% unconditionally, on the paper’s own arithmetic) — yet it circulates as “72% of studio ventures reach seed/Series A,” a mutation we received in our own briefing materials. Second, in 2020 most studio portfolios were too young to have failed; the surviving companies of self-selected member studios are not a base rate. And note what even the honest version measures: fundraising. Not revenue.

“Studio startups reach Series A in 25.2 months, versus 56 months for traditional startups”

Attributed to the same venture-studio association, and by some academic papers to later authors

Arithmetic artifact

An arithmetic artifact quoted as a measurement. 25.2 is the sum of two separate averages in the association’s paper (10.7 months zero-to-seed + 14.5 months seed-to-A); 56 is the same sum for traditional startups (36 + 20). Summing averages of different surviving cohorts is not a measured time-to-Series-A for any actual company. The one independent re-measurement we found (Max Pog’s 2023 research, with a public dataset) got ~33 months for studio startups versus ~56 traditional — still a real advantage, but 30% slower than the marketed figure. The provenance is now so muddled that a 2025 journal article attributes the 25.2 figure to a 2023 author. Still measures fundraising, not revenue.

“A well-tested corporate venturing approach succeeds 66% of the time”

A major strategy consultancy, 2022

Vendor-graded homework

The figure is real and the consultancy publishes it — but it is the consultancy measuring the venturing model pioneered by its own digital-ventures unit, on engagements it selected, against a definition of “success” that the public article does not disclose and that is not an external-revenue standard. It is quoted across the industry as if it were an independent base rate for corporate venture building. It is a Grade B portfolio claim from the seller of the model. (A rival consultancy’s equivalent numbers, for what it’s worth, come from a disclosed-method survey of 1,176 executives — and are far less flattering.)

“US companies face a $9 trillion shareholder-value deficit”

A major strategy consultancy, in sponsored business-press content, April 2025

True, but not what it sounds like

The figure is not a measured loss. It is the gap between current market capitalisation (S&P 500 minus the Magnificent Seven, banks, and REITs) and the value those companies would need to deliver top-tier 15%-annual TSR through 2029 — an aspiration gap, constructed by the consultancy’s own analysis, in a sponsored article whose remedy is that consultancy’s venture building. Quoted in decks as “$9T of value waiting to be unlocked by venture building,” it has quietly become a market-size claim. It isn’t one. (The same article, to its credit, is a genuinely useful archetype framework — which ends, as the whole genre does, at launch.)

The meta-point: all four of these travel because they’re the only numbers available. This field measures what’s easy (rounds, launches, model claims) and not what matters (external revenue). That gap isn’t just a citation problem — it is the stall.

Building or backing a corporate venture? I’d like to hear where it is.

Whether you’re a corporate with a venture that launched and hasn’t found its second customer, or a studio whose builds keep stopping at the parent’s front door — this is the transition I work on. I’ve led three corporate spin-outs through exactly this stage, and I sit on both sides of the table across eleven APAC markets.

If you’re building ventures with or for corporates, reach out. Happy to compare notes even if there’s nothing to sell.

Book a 30-min conversation Connect on LinkedIn

Method & honesty note

How this was put together

Every statistic on this page was traced back toward its primary source and graded: A = peer-reviewed study or disclosed-sample primary research · B = industry-published or self-reported data, sample not independently verified · EST = reasoned estimate. Where a number could not be traced, it is labelled as such or was cut. Body text describes source categories rather than naming organisations; the full named citations sit in the appendix below.

Three limits worth stating plainly. First, the global-survey figures — the best data in this field — are still executives self-reporting on their own ventures, GDP-weighted; “scaled” and “success” are judgment calls by the people who funded the ventures. Second, no published dataset cleanly separates parent-as-first-customer ventures from external-market ventures, so the causal mechanism at the centre of this page is an argument from the structure of the evidence and from operating experience — not a controlled comparison. It is stated as the falsifiable claim it is. Third, the venture-studio numbers rest heavily on one 2020 self-reported survey that the industry has been re-quoting for six years; the independent census work that exists (Pog, 2023–24) is one researcher’s disclosed-method effort, not peer-reviewed research. Where a number is soft, it is labelled soft rather than rounded into confidence.

The full limits file — seven notes, unabridged

The core mechanism is argued, not experimentally shown. No published dataset isolates parent-as-first-customer ventures and compares their external-revenue outcomes against external-market-first ventures. The page’s central claim is built from (a) the metric-ladder pattern in the published numbers, (b) the structural incentives of programmes and studios as publicly documented, and (c) operating experience from three corporate spin-outs. It is presented as falsifiable, and the falsification condition is stated on the page.
The best data is self-reported. The global consultancy survey series is the only longitudinal dataset in the field, but “scaled,” “success,” and revenue bands are all executive self-assessments, GDP-weighted. Grade A for method; softer than it looks for meaning.
The studio association’s paper could not be independently re-verified. The 2020 white paper is no longer hosted by the association (since rebranded); sample descriptions vary across secondary sources (commonly 258 startups). Our audit of it relies on the paper’s own published claims plus Pog’s independent census — one researcher’s disclosed-method work, not peer review.
Studio closure counting is inherently fuzzy. “Prolonged website inactivity” as a closure proxy may miscount in both directions.
Survivorship and selection bias run through everything. Studios publicise portfolios that raised; corporates publicise ventures that launched; nobody publishes “our venture booked eighteen months of parent revenue and then died.” The absence of captive-revenue post-mortems is itself consistent with the page’s thesis, but absence of evidence is weak evidence.
Consultancy sources are used for what they admit, not what they claim. The archetype frameworks, the 66% model claim, and the GenAI build-compression thesis are all Grade B vendor material; the page uses them as documentation of how the industry frames and incentivises the build phase — not as performance evidence.
Research constraints: web research only, no LinkedIn, conducted 2026-08-06 from a US-search vantage. A commercial research firm’s full CVC dataset and the studio association’s 2022 data report are paywalled/private and are noted, not quoted.

Appendix

Every number on this page, with source, grade and provenance

Sources appendix — 20 entries, plus the numbers we cut

Named sources appear here so every number can be checked; the body text deliberately describes categories rather than singling out organisations.

1 · Only ~22% of new businesses launched in the past ten years have successfully scaled Grade A
Leap by McKinsey press release, Dec 2021

McKinsey Global Survey on new-business building (annual, disclosed method); self-reported by executives.

2 · More than 4 in 5 new businesses not achieving US$50M annual revenue 4+ years after launch Grade A
Same Leap by McKinsey release (Dec 2021), GlobeNewswire mirror

Same survey series; self-reported revenue bands.

3 · 2024 McKinsey survey: n=1,176 senior managers & C-suite, fielded 21 May–2 Jul 2024; GDP-weighted Grade A
McKinsey, “How CEOs are turning corporate venture building into outsize growth,” Oct 2024

Fifth annual survey; method disclosed in “About the research” sidebar. Verified directly on page.

4 · ~Two-thirds of CEOs expect to build new ventures in the coming year; half call it a top-3 priority Grade A
Same McKinsey Oct 2024 survey

Verified directly on page.

5 · Novice builders 3× more likely to build a venture without product-market fit; experts ~2× success rate; 12× fifth-year revenue vs novices Grade A
Same McKinsey Oct 2024 survey

“Success” = meets/exceeds expectations for scale and growth (self-assessed).

6 · Ring-fenced investment → odds of new-venture success 28 percentage points higher Grade A
McKinsey 2021 findings, cited in footnote 11 of the Oct 2024 article

Earlier wave of the same survey series.

7 · Commercial roles among most-reported talent gaps; experts give ventures more decision independence Grade A
Same McKinsey Oct 2024 survey

Verified directly on page.

8 · 154 venture studios had ceased operations as of Sept 2024 — cumulative, against ~1,107 ever established; 2024 projected (not confirmed) as the first net-decline year on Q3 data, 17 registrations against 20 closures Grade B
Big Venture Studio Research 2024 (Max Pog)

Independent census; closure = announcement or prolonged website inactivity (may overcount quiet survivors, undercount zombie studios).

9 · GSSN: 84% of studio startups raise seed; 72% of those reach Series A (vs 42% traditional); 10.7 mo zero→seed, 14.5 mo seed→A (vs 36 + 20 traditional) Grade B
GSSN white paper “Disrupting the Venture Landscape” (2020), now hosted by Morrow · PDF mirror

Self-reported survey of GSSN member studios’ own portfolios; sample widely described as 258 studio startups; no independent verification; conditional structure of the 72% routinely dropped in re-quotation.

10 · Independent re-measurement: studio startups ~2.75 yrs (≈33 mo) to Series A vs 4.68 yrs (≈56 mo) traditional Grade B
Big Startup Studios Research 2023 (Max Pog), public dataset linked in-page

One researcher, disclosed dataset; corroborates studio speed advantage at a smaller magnitude than GSSN-derived 25.2-month figure.

11 · “25.2 months” attribution drift into academic literature Grade B
Journal of Management & Sustainability comparative paper (2025)

Attributes the figure to a 2023 author; illustrates provenance muddle. Used only as evidence of citation drift.

12 · BCG: model “achieves success 66% of the time”; 2–3× more likely to succeed than other approaches Grade B
BCG, “A Proven Model for Corporate Venturing” (2022) · PDF

BCG measuring the BCG Digital Ventures model; success definition and sample not disclosed in public article. “66%” wording verified on page.

13 · EY-Parthenon: $9T shareholder-value deficit; 70+ ventures built; four archetypes; “ready first customer within its own corporation” as Corporate Launchpad endowment Grade B
HBR sponsored content, 17 Apr 2025 · EY mirror

Sponsor content; $9T is EY-Parthenon analysis of S&P 500 ex-Mag-7/banks/REITs vs 15% TSR through 2029. All wording verified directly on the HBR page.

14 · EDB CVL 3.0: S$32M over two years; two tracks; up to 50% co-funding; sprint = concept validation; OI support “up till the pre-pilot stage”; ventures must HQ in Singapore Grade A
EDB media release · programme page

Government programme documentation; scope language verified directly on the EDB page (last updated Jul 2025).

15 · CVL 3.0 appointed partners incl. EY-Parthenon, BCG X, Bain, Wright Partners (w/ MING Labs), McKinsey, Stryber, Rainmaking, FutureLabs Grade A
EDB partner info pack (EY-Parthenon) · Wright Partners CVL page

Nine partners across the two tracks per EDB; named partners confirmed via EDB and partner pages.

16 · Corporate venture/CVC unit closures 2023–25: SAP.iO (Mar 2024), Verizon Ventures, Anglo American Decarbonisation Ventures; earlier: AmFam’s Tenney 110, BP Launchpad, General Mills G-Works Grade B
Global Venturing

Trade-press compilation; closure reasons attributed mainly to parent restructuring.

17 · Parent company frequently acts as spin-out’s first customer / resource provider (qualitative) Grade B
International Review of Economics (2023)

Academic case evidence on parent influence; does not quantify revenue outcomes of captive vs external-market ventures.

18 · BCG X 2025: GenAI compresses venture build cycles Grade B
BCG, “How GenAI Can Prompt a Faster, Less Risky Future of Venture Building” (2025) — bcg.com (via prior ArcShift research batch, 2026-08-06)

Vendor thought-leadership; used for the “building got cheap” framing only.

19 · Wright Partners builds for equity, not advisory fees; 16 ventures for corporates/family businesses Grade B
wright.partners · wright.partners/cvl

Vendor self-description; used for incentive-structure argument, not as performance data.

20 · “0 published stats measure third-party revenue” EST
Original finding, Arc Shift audit 2026

Bounded claim: of the statistics audited for this page (items 1–13 above), none measures external revenue with a disclosed sample. Falsifiable by counter-example.

Numbers considered and cut (couldn’t trace or too soft to use)

“72% of studio ventures reach seed” — the version received in briefing materials; a mutation of source #9 (original: 84% reach seed). Documented in the audit rather than used.
“60% of studio startups reach Series A” — derivable from #9’s own arithmetic (84% × 72%) but never independently measured; not used as a standalone figure.
CVL 1.0/2.0 cumulative venture counts — no consolidated official figures found on EDB public pages during this research; cut rather than estimated.
CB Insights CVC-inactivity counts — headline claims circulate but the underlying report is paywalled; Global Venturing’s named-unit reporting (#16) used instead.
“Two-thirds of venture studios are dead” (Venture Studio Index) — found but not source-checked within this cycle; excluded.

The build isn’t the hard part anymore.The first external dollar is.