Walk into any ministry of education in Sub-Saharan Africa and you will find articulate, well-meaning professionals with ambitious policy documents. Walk into a classroom in a peri-urban school in the same country and you will find something different: a teacher managing sixty students with chalk and a board, a curriculum designed for an economy that no longer exists, and an assessment system that measures memorisation rather than capability.
This is not a failure of intent. It is a failure of system alignment.
The Three-Layer Gap
Learning systems operate across three interdependent layers: policy design, institutional delivery, and classroom practice. When these three layers reinforce each other, the system produces learning. When they diverge, no amount of resources or effort closes the gap.
The problem is that they almost always diverge. And they diverge not because of corruption or incompetence, but because each layer operates on different timescales, incentive structures, and feedback mechanisms.
Systems fail not because the people inside them are failing. They fail because the architecture creates misaligned incentives, broken feedback loops, and no mechanism for the layers to correct each other.
What Misalignment Looks Like in Practice
A policy mandates continuous assessment. The institution trains teachers in the new approach. But the examination system, which determines student progression and school reputation, still rewards rote memorisation. Teachers, responding rationally to the incentive structure they actually operate within, revert to what the exam rewards. The policy fails. Not because of bad implementation. Because the system's feedback mechanisms pointed in the opposite direction.
The Systems Design Response
The question is not "how do we improve teaching?" It is "what does the system need to change so that teaching improvement is the rational, self-reinforcing outcome?" That is a fundamentally different question. And it requires a fundamentally different approach, one that starts with system diagnosis rather than program design.
The misalignment problem is solvable. But only if we stop treating it as a program design challenge and start treating it as a systems architecture challenge.