What a review of 91 food safety interventions reveals about how Africa measures success
That question should sit at the centre of every food safety programme. Yet many interventions end with a familiar set of proof points: the number of people trained, certificates issued, knowledge scores improved and workshops completed. These figures show that an activity happened. They do not, on their own, show that contamination fell, unsafe practices changed or fewer people became ill.
A new scoping review brings this measurement problem into sharp focus. Researchers examined 91 primary studies of food safety interventions implemented across Africa between 2000 and 2024. The review does not conclude that training is ineffective. Its more important message is that we often claim success before measuring the outcome that matters most: safer food.
A reassuring number, until you ask what was measured
At first glance, the results look encouraging. Seventy-two per cent of the interventions were reported as successful and a further 19 per cent as partially successful. But ‘success’ did not mean the same thing across the studies. In some, it meant a measurable reduction in a biological or chemical hazard. In others, it meant that participants knew more, reported better practices or were more willing to pay for safer food.
The measurement gap
44% measured hazard presence or levels42% measured knowledge, attitudes or practices
9% measured health outcomes
Those measures are not worthless. Knowledge can precede behaviour change, and hazard testing can be a strong indicator of risk reduction. The problem begins when an intermediate result is treated as the final result. A higher post-training score tells us that someone learned the correct answer. It does not tell us whether they could apply it during a busy shift, whether the facility supplied what they needed, or whether the improvement lasted.
Why before-and-after is not always enough
How an intervention is evaluated matters as much as what is measured. Forty-four per cent of the reviewed studies used a before-and-after design, while only 15 per cent used a randomised controlled design. A simple before-and-after comparison can detect change, but it cannot always tell us why the change happened. Seasonality, market conditions, a new manager, increased inspection or changes in water access may influence the result.
This does not mean every food safety project needs an expensive trial. It means the evaluation should be strong enough for the claim being made. If a programme says it reduced contamination, the evidence should include objective hazard data. If it says behaviour changed, observation at the point of work is stronger than self-report alone. If the intervention is expected to last, follow-up should continue beyond the final workshop.
Training cannot carry the whole system
Training and information-sharing were the most common intervention approach, appearing in more than half of the studies. That is understandable. Training is visible, scalable and comparatively easy to deliver. It is also essential when people have not received practical, role-specific instruction.
But knowledge operates inside a system. A food handler may know the correct storage temperature and still have no reliable refrigeration. A processor may understand cross-contamination but work without enough water, cleaning chemicals or separation between raw and finished product. A market vendor may know that food should be covered while facing heat, dust, limited storage and pressure to minimise losses.
Only about one in five interventions in the review identified a clear enabling environment. That gap matters. When equipment, infrastructure, supervision, incentives and enforcement are weak, training asks individuals to compensate for failures beyond their control. This is not an excuse for unsafe behaviour. It is a design problem.
When compliance becomes ritual
One of the studies included in the review described ‘ritual standards’ in the Lake Victoria fish value chain in Uganda. Methods associated with European Union compliance were performed, yet weaknesses in traceability, transport and the handling of substandard lots persisted. The visible act of compliance had become disconnected from the control it was supposed to provide.
The lesson extends beyond that case. A signed checklist is not the same as a controlled hazard. A certificate is not the same as safe production. An audit pass is not the same as consistent practice between audits. When the indicator becomes the goal, organisations can become very good at producing evidence of compliance without improving the underlying condition.
Technology helps when the operating reality supports it
The review also covered interventions such as aflatoxin biocontrol, improved cleaning systems and tools for better post-harvest handling. Technology can remove practical barriers that information alone cannot. But a tool is not automatically a solution. It must be affordable, maintainable, available when needed and compatible with the way people actually work.
The evidence illustrates both possibilities. In one intervention, training combined with free maize-drying tarpaulins improved post-harvest practices more than training alone. In another, a hygiene toolkit underperformed training alone after items were depleted or damaged. The difference is not ‘technology versus training’. It is whether the full intervention can function after the project team leaves.
The question is not whether to train
Training remains necessary. People cannot follow procedures they have never been taught, and clear communication is fundamental to food safety culture. The mistake is expecting training to carry an entire control system. The strongest interventions are likely to connect knowledge with the tools, processes, infrastructure, management support and regulatory conditions required to act on it.
The review also deserves careful reading. It was a scoping review, not a formal test of which intervention works best. The studies were diverse, the search was limited to English-language literature, grey literature was excluded and a formal risk-of-bias assessment was not used to decide inclusion. Positive results may also be easier to publish than failed interventions. The reported success rate should therefore be read as a map of the available evidence, not a final verdict.
We should still count the people trained. We just should not stop there. The more important count begins after the workshop: fewer unsafe practices, lower hazard levels, stronger operating systems and fewer people made ill by food.
Before calling the next intervention successful, we should be able to answer one direct question: what evidence shows that the food actually became safer?
Source
Pal, H. et al. (2026). “Food safety interventions in Africa: a scoping review of strategies, incentives, and enabling environments.” Frontiers in Sustainable Food Systems, published 27 July 2026.