An Honest Appraisal

This note is addressed directly to the readers of the 2022 publication. It is not part of the analytical argument. It is an accountability document.

Four years on, the question owed to readers - and to ourselves - is simple: did any of it matter?

The honest answer is: some of it did, some of it did not, and some of the world moved in directions that were neither predicted nor recommended.

The call for case studies was answered: six scaling case studies were published through the EADC programme. The call for data infrastructure was answered: the EADC built open data platforms in all three markets. Elements of the Evidence Public Dialogue Series were delivered through the Ecosystem Provocations series. The GESI: High-Growth Company Founding research provided the diagnostic foundation for addressing the gender financing gap.

The pan-African ScaleUp Institute was never established  the institutional gap that the African Scaleup Lab now directly addresses. No pan-African mentorship strategy was designed. No online scale-up academy achieved meaningful reach. The innovation leadership academy remains unbuilt. The syndication network model was never piloted. Mental health support for founders remains absent from the ecosystem's infrastructure. These are not minor omissions. They represent the difference between the ecosystem that exists and the one that is needed.

The pattern of stalled recommendations is not random. Each maps to a specific structural dynamic the loop analysis identifies - and the fact that these dynamics were not named in 2022 is itself part of the explanation for why the recommendations failed.

The pan-African ScaleUp Institute was not built because no actor with the structural mandate to convene it had the institutional incentive to do so. The actors best positioned to host it - the African Development Bank, AUDA-NEPAD, major bilateral donors - all operate inside Loop 1 (Programme-Rich, Capability-Thin) and Loop 2 (Role Collapse Amplification). An independent institution that measures programme effectiveness and publishes the results publicly is structurally threatening to every actor whose programme it would evaluate. The recommendation was sound. The political economy of its non-implementation was entirely predictable - and was not predicted.

The online scale-up academy and the mentorship network were not built because Loop 1 reliably directs resources toward activities that produce visible, countable outputs - cohort participants, events held, reports filed. An online academy produces diffuse, long-cycle, hard-to-attribute outcomes. A mentorship network requires sustained relationship investment with no quarterly deliverable. Neither generates the output metrics that donor reporting systems count. Both were therefore systematically underresourced relative to their potential impact.

The mental health infrastructure was not built because it generates no programme deliverable that any institutional reporting system measures. It is not a line item in any DFI logframe. It is not an output any accelerator can claim in its annual report. It is the recommendation that most directly serves founders and least directly serves the institutions nominally supporting them - and that structural reality is precisely why it remained unimplemented.

The syndication network model was not piloted because it would have required actors who currently compete for deal flow and programme mandates to collaborate around shared infrastructure. Loop 2 - Role Collapse Amplification - explains why this does not happen: bundled advantage is preserved by maintaining separate institutional positions, not by pooling them.

The pattern is consistent. The recommendations that were implemented - the case study programme, the data collaborative, the country-level ecosystem reports - were precisely those that produced outputs legible to programme funders within standard grant cycles. 

The recommendations that were not implemented were precisely those requiring institutional change, long time horizons, or the voluntary reduction of incumbent advantage. This is the loop operating as predicted. The 2026 recommendations are designed against this pattern rather than despite it - which is why each one specifies not just what should happen but who has to decide, what structural incentive they have to act, and what accountability exists if they do not.

The speed and severity of the capital correction was underestimated. The offshore incorporation dynamic was underestimated - the structural pressures driving African founders to incorporate in Delaware or the Cayman Islands are deeper and more systemic than the original analysis recognised, as documented in Offshoring African Startups. The distorting effect of multilateral and UN programme competition on local ecosystem markets was underestimated - a dynamic subsequently documented in When Multilaterals Compete. The pace at which AI would become a material factor in the scaling equation was underestimated: the original publication was completed before ChatGPT launched. Three years later, AI has restructured competitive advantage in almost every domain that African scaling ventures operate in.

TechCabal Insights' 2024 analysis documents that African startups pursued aggressive expansion strategies in 2024, entering 38 new markets - more than double the 16 recorded in 2023. M&A activity rose substantially, with approximately 35–39 transactions recorded versus around 29 in 2023. Venture debt and structured capital grew to 41 percent of total capital in 2025, according to Partech. African domestic investors moved from 23 percent to 45 percent of total venture fund commitments, according to AVCA - a genuine structural shift, though one led primarily by African DFIs and corporates rather than commercial pension funds or family offices, as the LumiBrief analysis of the same data notes. Former successful founders became angel investors and seed-stage backers at an accelerating rate.

These were not outcomes that the original recommendations produced. They were ecosystem adaptations to real conditions - the kind of endogenous response that the original analysis argued the ecosystem was capable of, but that required real stress to activate.