Safety models, measures and incentives...

Why do fatalities still occur?

Despite improvements in injury rates, the occurrence of fatal and catastrophic safety events is still an all too common occurrence.

In the post incident analysis with the benefit of hindsight, matters in relation to leadership, culture, production and profit motives get explored and assigned as contributing factors with actions to address. Yet the challenge remains on how to best try and predict causation ahead of time to prevent these events from occurring.

It is in this view that conventional safety approaches, measures and incentive might not be driving the results we had hoped for and the need for a course correction and change may be required. 

This leads us as a profession to challenge the limitations in our current methods and ask ourselves questions that might lead us to safer outcomes such as:

• Is the design of short and long term incentives and other targets really aligned to delivering safe outcomes?
• Has too much faith been placed in ‘safety as a value’ and not enough on the design of decision making environment to support safety remaining a priority when faced with other pressures?
• Do current safety management approaches and metrics adequately reflect the complexity of modern organisations and incident causation?
• Does an anchoring to aging safety models and metrics misguide management attention and mask reality?
• Do we adequately assess risk and monitor control effectiveness and procedural drift?

In this article we’ll explore contemporary theories and propose how we may better align our organisations to achieve safe outcomes.

Talk to us

To get a pdf of this article or have a deeper discussion about these issues, and to understand our health and safety capabilities, please contact:

Chris Alderson

Director Health and Safety Consulting

021 473 757

[email protected]

This publication has been prepared for general guidance on matters of interest only, and does not constitute professional advice. You should not act upon the information contained in this publication without obtaining specific professional advice. No representation or warranty (express or implied) is given as to the accuracy or completeness of the information contained in this publication, and, to the extent permitted by law, PricewaterhouseCoopers, its members, employees and agents do not accept or assume any liability, responsibility or duty of care for any consequences of you or anyone else acting, or refraining to act, in reliance on the information contained in this publication or for any decision based on it.

? 2017 PricewaterhouseCoopers. All rights reserved. In this document, “PwC” refers to PricewaterhouseCoopers which is a member firm of PricewaterhouseCoopers International Limited, each member firm of which is a separate legal entity.

Safety must be a value. But it will only be a priority if the business environment supports it. 

What drives the decisions we make? 

Normally the focus on understanding decision making in relation to safety occurs at the pointy end. That is, the injured worker or supervisor and why did they do what they did; despite their training, experience and understanding of the risk involved?

This is obviously important to understand. But if we subscribe to the notion that decisions senior management teams make impact safety, then understanding decisions at this level is also critically important. This can assist in the design of incentives programs, targets and a decision making environment that is better aligned to achieving safe outcomes.

Individual decision making is not always rational or ethically so, its is bounded and strongly dominated by the situational over the dispositional (personal attributes).

That is, external influences that stem from the environment such as immediate priorities and pressures are stronger drivers of emotions, decision making and behaviours. 

The consequences from this behavior (whether positive or negative) and how soon they are realised, influence how likely the behavior is going to be repeated and reinforces it.

Findings from our own research ‘Making executive pay work – The psychology of incentives’ supports similar claims, that reward design tends to assume that people make rational decisions and the assumption that performance-related pay works.

The characteristics of long term incentive programs such as high risk, complex and ambiguous performance conditions, arbitrary outcomes and multiyear deferral – suggest that individuals will discount them to a fraction of their economic value.

Researchers have long known that consequences lose influence on behaviour and individuals devalue outcomes as the consequence delay increases. The perception of risk also decreases as the negative outcome is delayed, perceived to have low probability and with a resultant increase in probability of unsafe behaviour occurring. 

Resultantly, this can create a self feeding cycle of structural and unrecognised control weakening that operates over an extended time period. For example, to meet immediate financial targets, direction is given to find cost savings in maintenance and service functions. Current safety measures, TRIFR and safety observations point to good performance and timely positive outcomes are achieved from the cost cutting exercise. All targets are met with no immediate negative safety outcomes, reinforcing that the correct decisions were made.

How senior management decisions contribute to major accident events has been highlighted in numerous investigations and case studies. Accident causation models have evolved from simple linear ‘cause and effect’ thinking to system and complexity models that try to more accurately reflect the nature and complexity of modern organisations that we’ll discuss next.  

Accident causation – from the simple to the complex

Searching for a root cause?

Accident causation models have evolved from simple linear ‘cause and effect’ thinking to theories on complexity, drift and systemic models to try and explain how accidents occur in modern organsiations.

Under these models, large and high risk organsiations are made up of many inter-related systems and components that make them complex. These organisations are not just formal, technical or engineering systems; they also have inherent managerial, behavioural and social dimensions that all interact at different organisational levels and time horizons to either contribute or take away from safety. 

Due to these systems being so intertwined, analysis needs to focus on the design of the system as a whole and not the single components e.g. not just the safety management system.

In these models, a single input can trigger a series of events that affect the numerous components, making direct cause and effect possibly untraceable. 

It is in this light that we can appreciate how seemingly reasonable decisions made at a distal senior management level to address an immediate need, can have long incubation periods that affect other components of the system resulting in unintended safety events.

These models can help us formulate better management approaches and measures that may yield greater safety benefits and challenge our thinking and anchoring we may have to familiar models and approaches. 

Understanding complexity requires more advanced analysis methods

A view put forward in Sidney Dekker’s book ‘Drift into Failure’ is that the growth of complexity in society has got ahead of our understanding of how complex systems work and fail. This concept can be applied to our understanding of how safety events occur in modern organisations.

Most organisations would have undertaken some form of analysis of their historical incidents to try and identify common causes and themes. 

These traditional approaches are limited by their inability to analyse large and disparate, but related, data sets relevant to safety events. 

Applying predictive analytic techniques to the safety domain can provide a level of analysis and predictive modelling capability far beyond these traditional approaches and capabilities of your typical health and safety function. 

Figure 1: Machine learning. One example of techniques used to explore safety incidents

By taking incident history data along with other data sources, these predictive modelling techniques can deliver longitudinal studies incorporating large data sets. These techniques can identify relationships and predictors of incidents before they occur that were previously not possible through traditional methods. We will discuss safety predictive analytics more later in this article. 

We need to improve our safety culture...

Safety culture is a hot topic. A ‘poor safety culture’ has been attributed as a factor in a number of catastrophic safety events. 

There is no shortage of organisations wanting to improve it and safety professionals and consultants claiming to be experts in it.

A typical definition of safety culture goes something like ‘the values, attitudes, behaviours and norms of an organisation in relation to safety’.

Many organisations have good safety values, attitudes and the standard systems one would expect for managing safety when audited. Yet post incident analysis reveals that behaviours and norms (the way they work) are not aligned to the espoused values and attitudes.

A limiting factors for some organisations in creating their desired safety culture is how they view the cultural change process and what’s required to make it stick. 

The cultural change process by some is viewed as a purely ‘hearts and minds’ exercise. But once the feel good factor of the aspirational visions and safety leadership course wear off, individuals are confronted with day to day operational realities. 

If there has been no change to the operational environment to support decision making and putting values into action, then divergence from the values and regression to ‘old ways’ occurs.

Leaders ultimately create the culture, they make safety important, demonstrate the values and drive operational discipline around process. But as discussed, as humans they are also fallible. So at all levels of the organisation, the design of the working environment needs to support safety values, decision making and priority setting.     

Hence, the integration of safety into key businesses processes at all levels of the organisations is important to normalise the ‘safe way’, as the way we work. This requires a holistic understanding of how businesses operate, their governance structures and their risks.

Safety Culture Vs Safety Climate

Safety climate exists at a more operational level and assesses the shared perceptions of the cohort on aspects such as managements adherence to safety values through decisions made and safety communication. It is a temporal state and dependent upon the current environment and conditions. Through regularly assessing safety climate, management teams can receive timely feedback on the perceptions and factors that might be driving behaviours. 

Anchoring to safety models and metrics that drive safety management

Beyond triangles, injury rates and safety observation targets

The limitations of measuring safety performance through lag injury rates and accident ratio triangles is well documented and founded.

Models such as Heinrich’s accident ratio triangle (first developed in 1931) have become synonymous with safety. This model had been accepted as a safety truism and has influenced how safety incidents are classified, reported, measured and how safety was managed.

It was incorrectly hypothesised that a direct ratio and predictive relationship existed between the incident classification types.

Therefore, by having less minor incidents, there will be less fatalities. The reality is that many minor incidents have no greater potential or causal relationship to fatal or other catastrophic / process safety related outcomes.

This model has been inappropriately used and expanded to try and explain more complex subjects such as risk, human behaviour and accident causation.

Catastrophic safety events such as Longford, BP Texas City, Macondo and Pike River provides us with some hard lessons learnt. They highlight the complexity of causal pathways for process safety incidents and the dangers of measuring safety through injury indicators and not monitoring the effectiveness of key safety systems and controls.

Monitoring and reporting on the performance of critical system and technical controls for managing the proximate causes to safety events is an area many organisations struggle to do well. Resultantly, senior management teams are not provided with the right information to inform decision making. 

Some reasons for this can be associated with;

• inadequate or bias risk analysis
• safety functions not providing adequate 2nd line of defence assurance to the business
• assurance and governance arrangements not aligned to risks
• inadequate integration of safety into business processes, reports and across organisational levels (vertical integration)
• focus on lag and non predictive lead indicators
• Management focus on compliance and certification versus risk management 

Possible next steps...

Building a case for change in the management approach to safety can be a difficult task. Especially in organisations that haven’t had a recent major incident and traditional indicators such as LTIFR that say safety performance is strong.

Barriers to change can also come from an organisations own safety function. With possible anchoring to familiar safety models and limitations in understanding of how businesses operate and how best integrate safety into key processes.

As safety functions are generally structured to ‘challenge the line’ and be the subject matter experts on safety, there can also be a lack of impartial challenge of the safety function itself in terms of its approach and activities.

In an attempt to better align our organisations to achieving safe outcomes; we put forward suggestions on:

• Understanding the “Critical Risk” zone
• Safety integration, operationalisation and monitoring
• Design of targets, lead indicators and incentives
• Advanced data analytic applications for safety 

Safety integration, operationalisation and monitoring

In complex organisations, a threat to safety can result from a lack of safety integration into key business processes and a lack of vertical integration across organisational levels.

Deficiencies in vertical integration can be attributed to weak reporting, communication and feedback processes across organisational levels and functions. This results in senior management teams not having suitable ‘line of sight’ on material safety risks that occur at an operational level to inform decision making.

A ‘stand alone’ safety management system that attempts to ‘influence’ or impose safety conditions on how the rest of the business operates, is not as effective as business processes that have incorporated safety into their original design. In turn, this helps to normalise safe behaviour, processes and supports the safety culture.

The management failures identified of the board and executive management at Pike River, in part can be attributed to deficiencies of safety integration. This relates to governance, risk, assurance and reporting processes that didn’t support in driving the organisational behaviours and activities required for delivering safe outcomes. 

Strong leadership is required to drive operating discipline around process. Even with this, procedural drift and migration will still occur at all levels over time.

Under pressure from financial, operational and psychological demands, work practices can drift from the documented process and can progressively weaken the system defences and contribute to events.

Hence having a robust assurance framework that tests the design and operationalisation of key system controls and monitors for drift, is important in helping to maintain the control environment.

Some suggested areas where organisation should consider formal integration of safety to support the decision making environment and to operationalise safety leadership, culture and risk management are:

• Board charter, reports and meeting agendas
• Risk policy and organisational wide risk framework that includes safety materiality criteria
• Audit / Risk committee charter, reports and meeting agendas
• Internal Audit Schedule
• STIP & LTIP schemes
• Remuneration Committee
• Business Conduct Policy
• Annual strategic and operational planning / budgeting / financial review processes
• Evaluation and approval processes for major capital projects / tollgates, mergers and acquisitions
• Operational CAPEX & project approval processes
• Internal / External Communications
• Operational meetings agendas and reports
• Supply / Procurement Processes
• Operational business continuity processes
• Change management processes
• Work management / planning processes for scheduling of maintenance, inspections, testing, control verification activities, audits, observations (includes activities performed by H&S function)
• Individual and Enterprise level KPIs targets (lead / lag)
• Leadership development training
• Recruitment & performance evaluation criteria
• On boarding / Induction processes
• Position descriptions

Safety incentive structure, lead and lag indicators

The following table outlines discussion points on possible safety incentive structures and indicators 

‘Big Data’ and Safety – Safety Risk Predictive Analytics 

The complexity of our organisations continues to increase through the further use of automation technology and the internet of things (IoT) that all serve to improve safety. However, they also pose challenges to the traditional ways we have been use to monitoring safety systems and analysing safety events.

We have discussed the limitations in monitoring an isolated systems in an increasingly interconnected and dynamic multi sociotechnical system organisation. The traditional approaches to measuring safety performance and analysis of incidents has failed to keep pace with technology and how businesses operate.

These traditional approaches are limited by their inability to analyse large and disparate data sets from multiple organisational systems that are relevant to safety performance and precursor to safety events.

Through applying predictive analytic and modelling techniques to the safety domain, we can provide a level of analysis and predictive modelling capability far beyond traditional approaches and capabilities of your typical health and safety function.  

By taking incident history data along with other data sources, these predictive modelling techniques can deliver longitudinal studies incorporating large data sets. These techniques can identify relationships and predictors of incidents before they occur that were previously not possible through traditional methods. This can lead to the develop of new leading indicators and more efficient allocation of resources for managing safety. 

Other data sources may include:

• Production & operational data – e.g. production / process rates, SCADA data such as record event log files of alarms, overrides, shutdowns etc.
• Maintenance records – e.g. maintenance scheduling performance, breakdowns / failures, inspections, tests, reliability data and performance.
• Financial data – e.g. spend on maintenance
• Human resources – e.g. training, rosters, working hours, workforce demographics
• Weather
• Geospatial
• Safety observations, audit findings etc.