Chapter 01
From finding to closeout
How a safety finding becomes a tracked observation with a named owner and a clock, and how the observation closes with evidence the audit will accept.
01
What counts as a finding
A finding is any tagged observation that the work in front of you does not match the standard. It might be an unsafe act caught on a walk, a near-miss reported by a banksman, a hazard logged during an audit, or a non-conforming control flagged by a captured view review. Findings come from three streams: people on site, scheduled audits, and the captured base itself. The discipline starts at the point of recording. Every finding is given a description in plain language, a location to package level, a severity grade, and a captured reference where one is available. A finding tagged to a 360 walk number, a fixed view timestamp, or a permit reference is much easier to route, defend, and close than one that lives only in someone's head. The most common failure here is undisciplined recording: a finding entered as "edge protection issue, north side" cannot be acted on by a different shift, and is hard to evidence weeks later. The team that takes thirty seconds at the time to add the package, the activity, and the captured reference saves itself an afternoon of reconstruction later. The standard the finding is measured against has to be visible too. A site that runs a written safety plan, with the controls named for each high-risk activity, can record findings as deviations from a known reference. A site that runs on intuition records findings as opinions, and opinions are hard to close. The first lesson of the loop is that the recording shape determines everything that follows. Get the fields right at the front and the loop has a chance of closing.
02
Routing the observation
A finding becomes an observation when it acquires a named owner and an expected response time. The owner is named as a role and a person — package manager for area B, HSE lead for the project — so the workflow continues to function when individuals rotate off. The response time is calibrated to severity: a major non-conformance gets hours, a routine housekeeping issue gets a working day, a strategic pattern gets a week. The escalation path is named ahead of time, not negotiated at the moment of breach. If the package manager has not responded within the window, the observation moves automatically to the project director with a short status note. If the project director has not responded, it appears in the weekly safety meeting at the top of the list. The captured base supports the routing by attaching the relevant clip or walk to the observation, so the receiver does not have to ask what they are looking at. The most common failure in routing is owner ambiguity: an observation tagged to "the contractor" or "site team" sits in nobody's queue and ages quietly until someone notices the breach. A clear role-and-person mapping resolves this. The second common failure is severity drift, where everything starts as urgent and the routing system loses meaning. A short severity rubric, agreed at the start of the project, keeps the queue legible.
03
Closeout evidence the audit will accept
Closeout requires evidence that the corrective action was actually taken. A typed note from the package manager saying "actioned" is not closeout. A captured view of the new edge protection installed, a follow-up walk that shows the lay-down area cleared, a signed permit rewrite, an attendance record from a refresher toolbox talk — those are closeout. The platform retains the open-to-close trail as a single record, so an auditor or a regulator can move from finding to evidence to closure in a few clicks rather than emailing five people for screenshots. A workflow that records findings without recording closures is theatre. It produces a number for the dashboard and a feeling of activity, but the next near-miss will reveal that nothing actually changed. The discipline is to treat the closeout artefact as part of the observation itself, not as a separate filing exercise.
04
Failure modes and what good looks like
The clearest failure modes are: findings logged without captured references, observations without named owners, closeouts without evidence, and a weekly meeting that does not look at the open list. Any one of those will collapse the loop. What good looks like is unromantic. The HSE lead can open the system on a Monday and see, for the previous week, the findings raised, the observations open, the observations closed with evidence, the ones that breached their clock, and the patterns by area. The package managers can see their own queue. The project director sees the escalations. The conversation in the safety meeting is short because the data is in front of everyone. That is the loop closing.
Practice
01. Take three findings logged on your project in the past fortnight. For each, write down the owner, the response clock, the closeout evidence, and the captured reference. Note which fields are missing.
Look for: A short table for three findings. Most projects will find at least one missing field per finding — typically the captured reference or the closeout evidence. The exercise reveals which part of the loop is weakest.
02. Draft a one-page severity rubric for your project: what counts as critical, major, minor, and routine, with an expected response time for each.
Look for: Four bands with response times and one or two example findings each. The rubric should fit on one side of A4 and be agreed with the project director before the next safety meeting.
Checkpoint
For a finding logged this week, can you show the owner, the clock, and the closeout evidence within two minutes?
Recommended reading