Resilient Supply Chain Podcast: Why Better Safety Metrics Still Fail to Prevent Serious Harm

The latest episode of the Resilient Supply Chain Podcast examines a persistent

contradiction in modern operations: safety metrics may improve while serious injuries

and fatalities remain stubbornly hard to reduce. Hosted by Tom Raftery, the episode

features John Dony, CEO and co-founder of the What Works Institute, and Mike Swain,

Technical Enablement Manager at Evotix. Together, they explore how traditional safety

measurement can obscure rather than reveal real risk, particularly in complex,

contractor-heavy environments. For supply chain leaders, the discussion matters

because it reframes safety as a test of governance, data quality and operational resilience

rather than a narrow compliance issue. The full episode is available at

www.resilientsupplychainpodcast.com.

When metrics stop reflecting reality

At the heart of the discussion is a simple but uncomfortable question: if safety numbers

are improving, why are the most serious outcomes not improving at the same rate?

Dony argues that many organisations still rely too heavily on metrics that are easy to

collect but harder to interpret. One of the clearest examples is the way increased

reporting is often treated as evidence that performance is getting worse, when it may

actually mean hidden risk is finally becoming visible. As he puts it, “sometimes it’s the

truth coming to light.” That is a difficult message for leadership teams used to

dashboards that reward falling incident counts, but it is essential if they want to

understand operational reality rather than just its presentation.

The problem is not that lagging indicators are useless. It is that they are often treated as

proxies for risk without enough context. Lost-time incidents and reportable events may

satisfy governance requirements or help with tender qualification, but they do not

necessarily tell leaders where the most serious exposure sits.

The supply chain dimension of safety risk

The episode makes clear that serious harm is rarely the result of a single isolated failure.

It tends to emerge across fragmented systems, shifting accountabilities and uneven

operating conditions, which is why the issue is so relevant for supply chain leaders.

Swain points to sectors such as construction and agriculture, where contractor-heavy

models, seasonal labour and changing site conditions create persistent risk. Large firms

may define safety expectations at the top of the chain, but smaller operators lower down

often lack the resources, capability or context to meet them in practice. That gap

between formal policy and real execution creates a familiar supply chain problem:

responsibility is pushed outward while exposure remains very much in place.

This is where safety becomes a broader governance issue. Resilience depends not only

on setting standards, but on understanding whether those standards can actually be

delivered across the network.

Why familiar models fall short

A major theme in the conversation is the limitation of the traditional safety triangle.

Both guests challenge the assumption that reducing minor incidents will automatically

reduce serious ones.

Swain notes that not all categories of harm sit on the same causal path. A large number

of low-level incidents does not necessarily predict a major fatal event. Dony adds that

many organisations have treated the triangle too prescriptively, building assumptions

around it that the underlying data never fully justified.

That matters because executive teams often prefer simplified frameworks. They are easy

to communicate and easy to benchmark. But simplicity can hide important distinctions.

Chronic exposures, ergonomic injuries and repetitive strain may not fit neatly into

conventional serious-injury models, yet they can still have life-altering consequences

and major operational implications.

Human factors and data gaps

The conversation also highlights how poorly many organisations still handle human

factors such as stress, fatigue and cognitive overload. These are widely recognised as

contributors to harm, yet they remain undermeasured and under-integrated into safety

systems.

Part of the problem is that these risks are harder to quantify. Part of it is concern over

privacy, liability and implementation. But the practical result is that many organisations

continue to focus on visible hazards while missing the deeper conditions that shape

decision-making and behaviour under pressure.

That challenge is closely tied to data quality. Swain describes how better systems

transformed reporting in one organisation, leading to an 800% increase in reported

issues. Far from being bad news, that allowed the business to see patterns, understand

causes and intervene more effectively. It also reinforced a broader point: more data only

helps if leaders are willing to treat it as a route to learning rather than a threat to optics.

The full episode is available at www.resilientsupplychainpodcast.com

AI as capability, not shortcut

The discussion of AI is notably pragmatic. Neither guest treats it as a cure-all. Instead,

AI is presented as potentially useful in areas such as predictive analytics, horizon

scanning and computer vision, but only when the underlying data and governance are

strong enough to support it.

Swain stresses that trust matters just as much as technical capability. If workers believe

AI is there to police them rather than protect them, adoption will fail. Dony’s summary

is even sharper: if organisations want AI to work, they need to “go back and get your

data right.”

That reflects a wider lesson across supply chains. Whether the subject is resilience,

procurement or safety, digital tools rarely fail because the concept is wrong. They fail

because the underlying data is incomplete, weak or disconnected from reality.

The strategic takeaway is straightforward. Safety cannot be treated as a narrow EHS

issue or a procurement checkbox. It is a test of whether an organisation’s governance,

operating model and data model are aligned with the conditions people actually work in.

As supply chains become more fragmented, scrutinised and pressured, that alignment

will become increasingly important to resilience, risk management and operational

execution.