Only 6% of executives say they are satisfied with their innovation outcomes, even though 84% call innovation critical to growth, according to McKinsey research cited in a Techpoint Africa analysis of corporate innovation programs. That gap, the analysis argues, is structural: most programs are designed to avoid real risk rather than to ship products.

The piece examines six programs at Google, Amazon, Microsoft, Unilever, Maersk and Mastercard, applying three hard criteria: a product used by real customers, a new business line, or a changed product-building process. For R&D managers, the common denominators are protected teams, metrics tied to shipped output, and insulation from internal politics.

Which programs actually shipped?

The six programs and their measured outputs:

• Google Area 120 (internal incubator): Aloud, Tables, GameSnacks — with roughly 60% of projects killed by design.
• Amazon Working Backwards (product system): AWS, Kindle, Prime.
• Microsoft Garage (intrapreneurship lab): Seeing AI, Soundscape, Microsoft Journal.
• Unilever Foundry (open innovation): 500+ startup pilots.
• Maersk Growth (venture studio): 40+ investments including Captain Peter, Huboo and Baton.
• Mastercard Start Path (equity-free accelerator): 500+ startups scaled across 55+ countries since 2014.

What does Google's incubator reveal about kill discipline?

Area 120 runs small teams of 3–5 people outside core business units, with a 6–12 month build window and a midpoint kill-or-scale decision. Projects needed early user traction — often 10,000+ users — a market of $10 million or more in TAM, and strategic alignment with Google's long-term bets. If a project missed those markers, it got cut.

The incubator is no longer standalone. Google reorganized it into its Labs division in 2021, and since late 2022 it has focused exclusively on AI-first projects aligned with corporate priorities — a reminder that even successful internal incubators are subject to portfolio repositioning.

How does Amazon's writing-first process filter ideas?

Amazon's Working Backwards requires a press release and FAQ written before any code. Six-page narrative memos, silent reading at the start of meetings, and debate only after shared context remove presentation theatrics. Ideas pass only on three criteria: a clearly articulated customer problem, economic viability at scale, and single-threaded leadership with end-to-end ownership. AWS, Kindle and Prime all emerged from this process.

Why do open models work at Unilever and Mastercard?

Unilever Foundry posts brand-defined challenges, runs 3–6 month startup pilots with clear KPIs, and routes exceptionally strategic pilots to Unilever Ventures for equity investment. Its focus has shifted toward AI-driven R&D and supply chain resilience.

Mastercard's Start Path takes no equity. It runs six-month cohorts of 10–15 startups with access to APIs, data rails and client relationships. Alumni have raised an estimated $25 billion-plus post-program, and Mastercard reports over 15,000 brokered connections between startups and its bank and merchant network — figures the company itself supplies and that managers should weigh as vendor-reported data.

Why do most programs fail?

The analysis identifies three failure modes:

• Zero-action innovation: hackathons and idea portals that generate applause but no budgets.
• Core-business red tape: legal, finance and business units blocking pilots that threaten existing revenue.
• Wrong metrics: counting ideas generated or participants engaged instead of products shipped.

On success rates, the piece puts 30–40% of initiatives in properly structured programs shipping something real, and only 10–15% of those becoming meaningful revenue drivers. It frames this as a portfolio, not a failure rate. On governance, it recommends a hybrid: centralized funding and protection, distributed execution near real customers.

The claim that separates winners, the author concludes, is not the model chosen — incubator, studio, intrapreneurship or open innovation — but whether leadership funds uncertainty, tolerates failure, and measures outcomes rather than activity.