Leading, measuring and implementing HOP: a capstone for Harrowgate

Two years after the line 3 event, Margaret Doyle asked Rachel Okoye to report to the board's safety and operations committee on what the HOP project had achieved and what it had not. Rachel asked Elena, Luis, Tanya, Sam and Grace to help her prepare. 'Not a celebration,' she said. 'An honest account. What changed, what did not, and what we would do differently.'

This final lesson draws on that account. It covers the leadership practices that HOP depends on, how Harrowgate Foods Ltd. measured whether HOP was working, and how it implemented the approach across three very different sites. It ends with a capstone: an event at the Dartmouth distribution centre in the project's second year, worked through using everything in this course.

What HOP asks of leaders

HOP is often described as a way of leading as much as a way of managing safety. Several leadership practices recurred in the Harrowgate Foods Ltd. experience.

• Ask more than tell. Edgar Schein's idea of humble inquiry, asking questions to which you do not already know the answer, out of genuine interest, was the model. 'What makes this job hard?' and 'What do you need?' replaced 'Are you following the procedure?'
• Go and see work as done. Leaders spent regular, scheduled time at the work, watching to learn rather than to catch, as Lesson 4 described.
• Respond well to bad news. The first words from Lesson 11 applied to leaders at every level. A leader who reacts badly to one report can undo months of trust.
• Remove obstacles. The most visible proof that leaders were listening was that problems raised by crews were fixed, and that people were told what had been done.
• Protect capacity. Leaders decided whether to add the second night technician, whether to keep six people on sanitation crews, and whether to move a restart time. Lesson 10 showed why those decisions mattered.
• Hold the line on critical controls. HOP is not permissive about the controls that prevent death. Leaders made clear that critical controls were not negotiable, while being open about everything else.

Middle managers were the hardest group to bring along. Supervisors and plant managers sat between executives asking for learning and production targets that had not changed. Some heard HOP as another demand on their time, and some heard it as criticism of how they had always managed. The project made progress only after it gave them something: a coach at each site, time released from paperwork that HOP made unnecessary, such as the old pre-job cards, and a seat in deciding which learning teams to run.

Marcus Bell's own change was one of the stories Rachel told the board. In the first year, he had defended the 25-minute changeover target. By the second, he had stopped reviewing changeover times at the morning meeting altogether and replaced the item with one question for each line lead: 'What got in your way yesterday?'

Measuring whether HOP is working

HSE-407 built a measurement framework around serious hazards and critical controls. The HOP project added to it, following the same rules: measure what matters, define it carefully, pair indicators that could be gamed, and act on thresholds rather than chase targets. The project avoided counting HOP activities for their own sake. The number of learning teams held, on its own, would have encouraged shallow sessions.

The measures fell into four groups.

• Learning: learning teams held on tasks with critical steps; share of actions completed on time; share of actions that changed the work rather than relying on training or reminders.
• Reporting and response: reports per 100 workers, paired with the share of reports that led to a change in the work; median days from report to first response; share of reports that named a condition rather than only a person.
• Critical controls and capacity: critical control verification pass rates and failed verifications found and fixed; capacity indicators from Lesson 10.
• Culture: pulse survey items such as 'I can tell my supervisor bad news without getting into trouble' and 'When I raise a problem, something is done about it', reported with the margins of error described in HSE-407.

Outcome measures remained, including recordable injuries, SIF-potential events and serious injuries, but the board was reminded each quarter of HSE-407's warning that two years of injury data at a company of Harrowgate Foods Ltd.'s size could not prove or disprove that HOP worked.

Nina produced the two-year summary for the board.

# Two-year summary of selected indicators, all sites
rows = [
    ("Reports per 100 workers per year",               48,   86),
    ("Share of reports leading to a work change (%)",  22,   47),
    ("Median days from report to first response",      19,    6),
    ("Learning team actions that changed the work (%)", None, 64),
    ("Lockout verification pass rate (%)",             81,   97),
    ("Failed critical control checks found and fixed", 38,  112),
    ("Critical roles with one trained person",         27,    9),
    ("Pulse: can tell supervisor bad news (% agree)",  54,   73),
    ("SIF-potential events reported",                  61,   94),
    ("Serious injuries (SIF)",                          2,    0),
]
print(f"{'indicator':50} {'baseline':>9} {'year 2':>7}")
for name, before, after in rows:
    b = "-" if before is None else before
    print(f"{name:50} {b:>9} {after:>7}")

indicator                                           baseline  year 2
Reports per 100 workers per year                          48      86
Share of reports leading to a work change (%)             22      47
Median days from report to first response                 19       6
Learning team actions that changed the work (%)            -      64
Lockout verification pass rate (%)                        81      97
Failed critical control checks found and fixed            38     112
Critical roles with one trained person                    27       9
Pulse: can tell supervisor bad news (% agree)             54      73
SIF-potential events reported                             61      94
Serious injuries (SIF)                                     2       0

Rachel explained to the board that several numbers had risen on purpose. More reports, more failed verifications found and more SIF-potential events reported were signs that people were seeing and saying more, not that the work had become more dangerous. The fall in median response time from 19 days to 6 was, in her view, the single best measure of whether leaders were listening. Two years without a serious injury was welcome but, as HSE-407 had shown, not yet statistically meaningful on its own.

Implementing across three sites

The three sites differed in size, workforce and history, and the project learned that HOP had to be implemented with each, not delivered to each.

The roadmap had four phases. In the preparation phase, the executive team and site leaders learned the principles, agreed in writing how the company would respond to failure, and chose two pilot areas: line 3 and the distribution centre's freezer operations. In the pilot phase, the pilots used learning teams, the new pre-job brief, critical control verification and post-job reviews, with coaching. In the spread phase, trained coaches at each site took the practices to other areas, procedures were rewritten to include critical steps, and the new measures were added to reports. The sustain phase, which began at about eighteen months, included an annual review of the approach, refresher sessions, protected time for leaders at the work, and regular sharing of stories at all three sites.

Several lessons from implementation were worth recording for other companies.

• Start with a real story. The line 3 event did more to persuade people than any training session.
• Change the response to failure early. Withdrawing Ben's warning, and saying so publicly, was the moment people began to believe.
• Fix things fast. Small, quick fixes from early learning teams built trust that larger changes needed.
• Involve the agencies and contractors. At the distribution centre, HOP stalled with agency workers until the agency supervisors were trained alongside Harrowgate Foods Ltd. supervisors.
• Embed in the existing system. HOP practices were written into the occupational health and safety program, the investigation procedure, management of change and procedure control, so they would outlast the people who introduced them.
• Expect setbacks. One plant manager reverted to discipline after a costly product loss, and the project had to rebuild trust in that area.

Capstone: an early departure at the docks

In the project's second year, at 11:40 p.m. on a Friday in November, a trailer at dock door 9 of the Dartmouth distribution centre pulled away while an agency forklift driver, Kwame Asante, was reversing out of it with a loaded pallet. The trailer moved about 1 m (3 ft) before the driver stopped. The forklift's rear wheels dropped off the edge of the dock plate, and the truck tipped backward against the dock frame. Kwame was wearing his seat belt and was shaken but unhurt. The event had clear SIF potential: forklift tip-overs and falls from docks are among the most common causes of forklift deaths.

Grace, Sam and Elena led the response. The steps below show how the course's ideas were applied.

The dock 9 event was the subject of the last story Rachel told the board. She noted that it had happened two years into the project, at the site with the most experience of learning teams, and that some would see it as evidence that HOP had failed. Her view was the opposite. The event was reported within minutes. Nobody tried to hide the out-of-service restraint or the wrong-door messages. The people involved were cared for and spoke freely. And within six weeks, conditions that had existed for years had been fixed at every dock in the company. 'Error is normal,' she said. 'This is what it looks like when the system learns from it.'

Looking back over the course

HSE-408 has followed Harrowgate Foods Ltd. from a near miss on a bakery line to a company-wide way of working. The ideas are not complicated: people make mistakes, context shapes what they do, blame fixes nothing, learning is vital, and the response to failure matters. What is difficult is acting on them consistently, especially under pressure, when the old habits of finding someone to fix are quicker and more familiar.

The practices in this course give those ideas a concrete form. Understanding performance modes, precursors and drift explains why errors happen. Error-tolerant design, usable procedures and critical controls make the work able to fail safely. Learning teams, pre-job briefs and operational learning show how work is really done. Capacity gives the margin to absorb surprises. A restorative response keeps people talking. And leaders who ask, listen, measure what matters and protect capacity keep the whole system moving.