Corporate Venture Building: A 90-Day Plan With Go/Kill Gates

The short answer: plan a corporate venture as three 30-day phases that take it from a blank page to a first release for real early adopters: choose what to test, test demand, build the first release. Each phase ends at a gate where a named sponsor decides to continue, change one thing or stop, against criteria written down before the clock starts. Ninety days is a planning target that produces evidence, not a business: McKinsey reported in 2021 that only 22 percent of new businesses launched in the past ten years had successfully scaled.

Disclosure: I sell interim and fractional CMO work, so I am not a neutral party here.

I have launched 53 ventures, and at BNP Paribas I validated a corporate venture and handed it back within four months. That is context, not proof that this plan works. Steve Blank, in his May 2013 Harvard Business Review article Why the Lean Start-Up Changes Everything, called the claim that the lean process makes individual start-ups more successful "too grandiose". His own claim was different: across a portfolio of start-ups, lean methods "will result in fewer failures than using traditional methods". I read this plan the same way. It will not rescue a weak idea, but it shows you early and cheaply which ideas are weak.

The plan on one page

Skip this plan if the idea is really a feature of an existing product, if nobody will own the gates, if the outcome is already decided, or if the concept depends on a license or a hardware build that cannot be tested inside 90 days. In the last case, start the license work first and test demand in parallel.

Before day 1: five decisions

The gates only work if these exist before the clock starts. McKinsey lists strong CEO sponsorship and funding that follows validation among the four things an incumbent needs to build businesses repeatedly.

1. A sponsor who holds the budget. One executive who can say stop and who will be in the room on days 30, 60 and 90.
2. Money released at gates. McKinsey describes incumbents earmarking capital and releasing it when a venture reaches agreed development targets. Fund phase 2 at gate 1 and phase 3 at gate 2.
3. A minimum prize. The smallest yearly revenue or profit, in euros, at which the parent would care. A concept that cannot reach it even if everything goes right is a feature or a side project.
4. Pass lines. One page per gate: what you believe, how you will test it, what you will measure, and the result that means you were right. Strategyzer's Test Card has those four fields and calls the last one the threshold. At each gate the venture brings a one-page memo: the pass line, the result and the decision it asks for.
5. A home for the venture. The person who runs it if it works, and the business that takes it if it is handed back, named before day 1.

Phase 1: Choose what to test (days 1 to 30)

The aim is three to five concepts that deserve a test, not a long list.

What to do. Spend weeks 1 and 2 talking to customers and to your own sales and service teams, before the workshop. In week 3, run the ideation workshop with the interview notes on the wall. In week 4, write each surviving concept on one page: the customer and the problem in their words, the solution, the riskiest assumption, the evidence that would settle it, and the parent asset that gives it an edge.

That last item matters. McKinsey lists talent, funds, market insights, intellectual property and data among the assets that give a corporate venture an edge over a stand-alone start-up, and names bureaucratic interference and under-use of the parent's assets as common reasons these efforts fall short. So give the venture its edge asset with few conditions, its own budget cycle and one person in the core business who opens doors.

Write the business case backwards. The same article describes a "reverse" profit and loss statement: start from the profit and work back through costs to revenue. Its examples of assumptions to test are customers won, conversion at each sales stage and the cost of acquiring customers. Phase 2 tests exactly those.

Output. Three to five one-page concepts, each with a named owner.

Gate 1, day 30. Decide which one or two concepts get a test budget: those whose riskiest assumption you can test in 30 days with customers you can reach. Each needs ten or more conversations with people who match the target customer. Ten is my rule of thumb, not a standard, but a concept with no customer conversations is an opinion.

Kill a concept if:

• it has no owner and the sponsor will not name one;
• you cannot name the parent asset that gives it an edge;
• the prize, with every assumption going right, is below the minimum prize;
• you cannot reach the target customers inside the window;
• a license, approval or legal question will outlast the 90 days and nothing can be tested around it.

Phase 2: Test demand (days 31 to 60)

The aim is to learn whether strangers want it, at a price the plan needs, before anything is built. Blank's exhibit "Listen to Customers" calls this stage customer validation: the venture tries to validate customers' interest "through early orders or product usage" and changes a hypothesis if there is none.

What to do. In week 1, build one page per surviving concept, with tracking. In weeks 2 and 3, buy traffic from the people the concept is for, on a fixed budget. In week 4, read the result against the pass line and write the gate memo. The page says plainly that the product is not on sale yet, shows the price you expect to charge, and asks for the most committing action you can honestly offer: a waitlist sign-up at minimum, a refundable deposit or pre-order where you can deliver or refund. The method is in smoke testing product ideas.

Write the pass line first. Illustration, not a client case: you can afford €150 to win a customer, you assume one in five sign-ups becomes a customer, and a click costs €1.50. A sign-up is then worth at most €30 (150 ÷ 5), so the page must turn at least 5 percent of clicks into sign-ups (1.50 ÷ 30). With €6,000 of media you buy 4,000 clicks and need 200 sign-ups. If the page converts at 3 percent you get 120 sign-ups at €50 each, which implies a €250 cost per customer (50 × 5). The venture has failed its own test, and the right response is a decision, not a longer test.

Output. Live pages with real traffic and a one-page result against the pass line.

Gate 2, day 60. Decide whether to build, change one hypothesis and retest once, or stop. Build only if the pass line is met on the most committing signal you offered. Change one hypothesis if the interest is real but sits on a different promise, price or segment from the one you planned, and retest once on a fixed budget. Do not extend a test until it passes: a test you keep running until it looks good proves nothing.

Kill the venture if:

• the pass line is missed, and the best variant misses it too;
• the cost per sign-up or deposit is above what the economics allow;
• the interest comes from people the parent cannot serve, or from staff, partners and friends;
• the price that attracts interest is below the price the plan needs;
• a retest has already failed once.

Phase 3: Build the first release (days 61 to 90)

The aim is the smallest release that lets real early adopters do the one job the page promised, and shows whether they come back and pay. It is not a scaled-down final product.

What to do. Days 61 to 65: a five-day design sprint. The sprint, developed at GV, runs Monday to Friday: map the problem, sketch competing solutions, decide, build a prototype, test it with real customers. It settles what the first release holds and what it leaves out. Days 66 to 75: build only that. Days 76 to 90: put it in front of early adopters, invited from the people who joined the waitlist or paid a deposit in phase 2, because they asked for it. Ten days buys a thin release: a service delivered by hand, a no-code product or a prototype wired to a real service. If the build needs longer, say so at gate 2 and move the dates. Do not squeeze the learning period.

The team. One accountable lead who controls spend, plus product and design, engineering, a go-to-market owner and one person from the parent who opens doors: four to six people, my suggestion rather than a standard. The go-to-market seat can be a manager borrowed from the parent, a permanent hire, or an interim or fractional lead. Test whoever fills it on one question: can they run a smoke test and an early-adopter launch without asking permission for every euro of media?

Output. A controlled release to a small group of early adopters, not a public launch.

Reading a small group. With 20 early adopters, one person is five points of any percentage. If 8 come back in week three, that is 40 percent, but the 95 percent confidence interval runs from 22 to 61 percent. A group that size tells you why people stay or leave, not the rate, so set gate 3 on behavior you can observe in people.

Gate 3, day 90. Decide whether to scale, extend once, hand back, or stop. Illustration of a pass line: at least 8 of the first 20 early adopters complete the core job in week one without help, at least 5 agree to pay or sign a letter of intent, and the acquisition cost you saw is below the level the plan allows. "Extend once" means up to 30 more days on one named hypothesis, agreed with the sponsor, never open-ended. "Hand back" means the venture goes to the business owner you named before day 1, with the system documented.

Kill the venture if:

• early adopters try it and do not return;
• they will not pay, or will pay less than the plan needs;
• the acquisition cost you saw is above the allowed level and you cannot name a credible way to bring it down;
• nobody in the parent will own it;
• the team cannot be staffed for the next phase.

What 90 days does not tell you

A first release shows that some people want the thing. It does not show that the business works. In the same 2021 analysis, two-thirds of the value created in new-business building is created in the scale-up phase. In a 2019 article on corporate business building, McKinsey puts time to profitability at usually three to five years, with capital needs that can approach or exceed $100 million per business. McKinsey sells business-building services, so weigh these figures as coming from an interested party.

So a pass at gate 3 does not license a scale-up. McKinsey's Scale or fail says to confirm three things first: demand validated on customer-oriented measures such as revenue or users, unit economics that are profitable or on a clear path to profit, and a leadership team able to steer rapid growth. If you want the first 90 days run for you, that is the venture builder work.

Frequently Asked Questions

What is the biggest challenge in corporate venture building?

McKinsey's survey of 93 executives at more than 70 companies led it to conclude that the single biggest challenge in scaling new ventures is getting talent and culture right. In the first 90 days I would add decision rights. If the venture needs the parent's approval for each euro of media or each change to a page, the clock is lost before the test starts, which is why the sponsor and the gates come before day 1.

How do you measure the success of a corporate venture?

By stage, against a pass line written before each test: customer conversations at gate 1, sign-ups or deposits at gate 2, and whether early adopters do the core job, come back and pay at gate 3. McKinsey lists soft measures of success among five common pitfalls and notes that funding should rise as milestones such as adoption and customer-acquisition cost are met. Revenue and market share come later.

How does agentic marketing fit into venture building?

It speeds up drafting and analysis, not the customer conversations. Agents can generate page and ad variants from a template, tag and summarize interview transcripts, and tidy the results table. They cannot sit in the interview, and a person makes the go or kill call at each gate. See what is agentic marketing for the wider picture.

Is 90 days realistic?

As a planning target for a software or service concept with no license dependency, yes. It is a target, not a promise. Hardware, licenses and regulated products take longer, and the plan says to move the dates at gate 2 rather than squeeze the learning period. A first release is also only the start: McKinsey puts time to profitability for corporate new businesses at usually three to five years.