Why food testing programmes need to get smarter, not just bigger

Food testing programmes rarely become complex overnight. They are built layer by layer.

A previous incident leaves a new test behind. A customer asks for additional reassurance. A regulation changes. A new supplier, ingredient or market introduces another point of uncertainty. Each decision may be sensible at the time – yet combined, those decisions can create programmes that are larger, more expensive and harder to challenge than the risk picture they were designed to manage.

Over time, a testing plan can begin to resemble an archive of past concerns as much as a map of current risk.

That matters because the risk environment facing food manufacturers is becoming more active and less predictable. The European Commission’s 2025 Alert and Cooperation Network annual report recorded 10,490 notifications – an 11 percent increase on 2024 – while Rapid Alert System for Food and Feed (RASFF) notifications rose by two percent and alert notifications increased by 14 percent. At the same time, investment in earlier warning systems, including a £650,000 FSA and BBSRC-backed research network focused on future food safety risks, points to a sector increasingly committed to anticipation alongside response.

For manufacturers, that puts the emphasis on responsiveness: testing programmes must evolve as evidence changes, rather than simply accumulate over time.

For Nia Owen-Cortez, Vice President Quality Assurance at Mondelez Europe, that starts with a shift in how businesses build evidence. Due diligence remains non-negotiable, but the sources of assurance are expanding.

“Due diligence will always be a necessity, and businesses must continue to meet their customers’ expectations,” she says. “However, where I see a meaningful shift is in where companies can source that data and in the growing need to build more probability-based testing programmes.”

A fixed schedule can provide consistency. Over time, though, it can also become blunt. A smarter programme draws on wider intelligence, understands probability and can explain why each test is still needed.

From internal results to wider intelligence

Internal testing data remains essential. It shows whether specifications are being met, whether controls are working and whether contaminants are being monitored.

It is also only one view.

In a globalised food system, the same hazards, commodities and sourcing regions often matter to many manufacturers at once. A residue trend, a recurring contaminant or a pattern linked to a particular origin may be difficult to spot inside one company’s dataset. Across a wider pool of results, however, those signals can become sharper.

“We can now draw on a wide range of external data sources, which means we no longer need to generate all due diligence data internally to demonstrate compliance,” Owen-Cortez explains. “For contaminants such as pesticides and heavy metals, for example, we can combine external information with our own results to build a far more complete picture.”

Pesticide residues show why that broader view matters. The European Food Safety Authority (EFSA)’s 2024 pesticide residue monitoring report analysed more than 125,000 food samples collected across Europe. In the EU-coordinated sampling programme, 98.8 percent of 9,842 samples were compliant with EU legislation, while national control programmes found 98.2 percent compliance across 86,449 samples. For manufacturers, that kind of data does more than confirm broad compliance. It helps show where exceptions occur, which products or origins require closer scrutiny and where monitoring must stay alert.

The ability to be agile – adjusting testing in response to a heightened risk – is essential. But agility works in both directions.”

Heavy metals illustrate the value of a longer view. EFSA notes that metals such as arsenic, cadmium, lead and mercury may enter food through natural occurrence, environmental contamination, farming, industry, processing or storage. They can also accumulate in the body over time, making them a long-term risk management issue rather than a one-off testing concern.

At Mondelez Europe, Owen-Cortez points to a trusted third-party model the business uses, in which testing data is shared with anonymised peers through a dedicated platform. Its value lies in scale. Thousands of additional data points beyond the company’s own programme can provide a broader view of general compliance for substances such as pesticides and heavy metals.

With that wider view, testing becomes less about isolated results and more about probability. Quality teams can ask where a contaminant is more likely to appear, whether that probability is changing and which materials justify closer attention.

“This broader dataset can then feed into models that quantify the probability of certain contaminants appearing in specific materials,” Owen-Cortez says. “In turn, that allows us to adapt our testing programme according to risk rather than relying on a fixed schedule. The result is a more fluid testing programme – one that enables us to pivot quickly and screen for emerging risks as they arise.”

A programme that can move with risk also needs the confidence to move in both directions.

When risk-based testing means doing less

Responsive testing is often associated with increased activity: a new screen, a higher frequency, a closer look at a supplier or commodity. The harder decision is usually the reverse.

Adding a test is easy to defend. It looks prudent. It shows action. Removing one can feel more exposed, even when the data shows a risk is well controlled.

Yet a programme that only grows can lose precision. As resources become tied to legacy activity, quality teams can spend too much time maintaining the schedule rather than asking whether it still reflects the current risk profile.

Owen-Cortez is clear that agility cannot only mean adding activity.

“The ability to be agile – adjusting testing in response to a heightened risk – is essential. But agility works in both directions.”

That is a crucial point for manufacturers under pressure to do more with finite resources. Testing is a verification activity. It can confirm whether controls are working but it does not replace supplier assurance, process control or well-executed prerequisite programmes.

Where supplier, manufacturing and supply chain data can be brought together, risk can be profiled more holistically. That gives quality teams a stronger basis for increasing scrutiny where uncertainty is greater and reducing testing where confidence is already supported by evidence.

Smarter testing is more deliberate testing. It requires the confidence to add, maintain or reduce activity according to what the evidence shows.

Governance turns flexibility into assurance

That confidence depends on governance. Without it, flexibility can look inconsistent. Without review, consistency can become stagnation.

Testing programmes carry history: a supplier issue that has been resolved, a legacy specification, a historical complaint or a past risk signal can remain embedded long after the original rationale has weakened. Unless someone formally challenges the programme, yesterday’s concern becomes tomorrow’s routine.

“Governance is key here,” Owen-Cortez says. “Alongside the data-driven approach outlined above, we must remain diligent in periodically reviewing our testing programmes as part of our internal controls.”

That review should do more than confirm that last year’s plan has been repeated. It should test the logic behind the programme. Has supplier performance changed? Are results leading to action? Is the test still linked to a defined hazard, customer requirement or compliance need? Does the current frequency reflect the probability of the risk?

The scale of that task explains why advanced analytics are becoming part of the food safety conversation. The European Commission’s TraceMap tool, introduced in March 2026 to help authorities connect food chain data and investigate potential risks, illustrates the growing role of advanced analytics in food safety oversight. For manufacturers, similar analytical capabilities could support more responsive testing programmes by identifying emerging patterns across suppliers, ingredients and sourcing regions.

AI and analytics therefore offer a way to make programme review faster, sharper and more evidence-led, while keeping expert judgement at the centre.

“What was once an arduous, resource-intensive task – reassessing where, when and what to test – will become significantly more efficient and insight-led,” Owen-Cortez says.

Technology only creates value when organisations are prepared to act on what it reveals. A dashboard does not improve assurance by itself. A model does not strengthen a programme unless its findings are reviewed, owned and translated into decisions.

The more data-led testing becomes, the more important it is to distinguish useful intelligence from information that simply accumulates.

“Organisations generate mountains of data every single day. If that data is not organised, cleansed and mined effectively, it becomes little more than ‘zombie’ data – present in abundance yet serving no purpose.”

For testing data to matter, it needs a route into decisions. A trend that is visible but not acted upon does not strengthen assurance. A result that never influences testing frequency, supplier oversight or risk profiling has limited value.

The most useful insight may even be that a test is no longer needed. That conclusion can be uncomfortable, but it is exactly what a mature, governed programme should be able to reach. In practice, that might mean easing a routine screen for a consistently controlled material while increasing scrutiny of a higher-risk supplier, commodity or origin where the evidence is shifting.

Those decisions also need to be understood beyond the quality function. Increasing, reducing or redirecting testing can affect suppliers, laboratory capacity, production schedules, lead times and customer commitments. Testing plans therefore need wider ownership across the business, with quality teams able to explain the rationale behind what is tested, when and why.

Organisations generate mountains of data every single day. If that data is not organised, cleansed and mined effectively, it becomes little more than ‘zombie’ data – present in abundance yet serving no purpose.”

Every test should have a reason

The next generation of food testing programmes will be judged by the quality of the thinking behind the schedule. Smarter programmes will draw on internal and external intelligence, use probability to guide verification, increase scrutiny when risk signals strengthen and reduce activity when evidence supports confidence.

The strongest testing programmes will be those that can learn, adapt and explain themselves, where every test still has a reason to exist.