Most organisations know their safety data could be better. They've seen the inconsistent records, the incomplete fields, the categories that no longer reflect how work is actually done. Many have launched initiatives to address these issues. Yet somehow, despite good intentions and genuine effort, the data doesn't improve.
The problem is rarely technical. It's structural, cultural, and often, invisible to leadership.
1. The day job always comes first
Safety data is created by people whose primary job is something else. Supervisors, operators, and frontline workers record incidents and observations alongside their core responsibilities. When time is short and production pressure is high, reporting suffers first.
This isn't a failing of individuals. It's a predictable outcome of how work is organised. Until organisations address the structural tension between production demands and reporting expectations, data quality initiatives will continue to struggle.
2. No one owns the problem
Data quality sits awkwardly between functions. IT manages the systems. Safety teams define the requirements. Frontline staff create the records. When quality falls short, everyone assumes someone else is responsible.
Successful improvement requires clear ownership, not just of the data, but of the conditions that enable good data to be created. Without that, improvement efforts become isolated projects that fade once initial enthusiasm wanes.
3. Incentives are misguided
Leaders say they want better safety data. But if managers are measured on production output, schedule adherence, and cost control, reporting becomes an afterthought. If raising concerns leads to scrutiny rather than support, people learn to stay quiet.
The gap between what organisations say they value and what behaviours they actually reward is one of the most persistent barriers to data quality. Closing it requires visible, consistent action, not just policy statements.
4. Process improvement isn't enough
Many organisations respond to data quality problems by redesigning forms or implementing new software. These changes can help, but they rarely address underlying issues. If the conditions that created poor data remain unchanged, like time pressure, misaligned incentives, and lack of follow-through, new processes will eventually produce the same results.
What actually works
Improving safety data means examining how data is created, who creates it, what pressures they face, and whether the information captured reflects operational reality. It means leadership commitment that goes beyond words and visibly acts on reported information, communicating changes back, and protecting time for quality reporting.
Want to learn more?
5 Questions to ask before investing in AI for QHSE
Risk Factor Episode 3 - One technology, two truths: The AI conversation we need to have
.webp)