This step records the basic facts and tells you how deep the investigation needs to go.

1. Fill in the case details: investigation title, organisation, date of incident, location and lead investigator. Type the date however you like (14/05/2026 or 14 May 2026). It converts to 2026-05-14 when you leave the field.
2. Set Actual severity. What harm occurred.
3. Set Potential severity. The credible worst case. This is the field most people get wrong. A dropped load that missed a worker by a metre had a potential of Catastrophic, whatever the actual outcome.

4. Write what happened. Facts only: who, what, where, when. Leave why for the analysis. If your summary already contains a cause, you have decided the answer before looking at the evidence.
5. Tick any complexity factors that apply: multiple parties, contractors or departments; similar prior incidents or near misses; plant, design or engineering failure; notifiable, or likely inquest or prosecution.
6. Read the Recommended level and click Switch to Level X if it differs from the current level.

How the recommendation works

• The tool takes the higher of actual and potential severity. Minor or Moderate starts at Level 1, Serious at Level 2, Major or Catastrophic at Level 3.
• Two or more complexity factors push it up one level.
• A notifiable event, or one likely to reach an inquest or prosecution, is never below Level 2.

You can override the recommendation. If you run below it, the check panel raises a red note, and that note is printed as unresolved when you reach the report.

Everything else in the tool links back to this register. A claim with no evidence ID behind it is an assumption, and the tool treats it that way.

1. Click Add evidence item. The tool assigns an ID (E01, E02 and so on).
2. Short title, for example “Maintenance log, conveyor 3, March to May”.
3. What it shows. The specific facts, not your interpretation of them.
4. Type: witness statement, document, CCTV / image, physical test, site inspection, data / record, incident report, expert report, legal record or other.
5. Source / reference. Where someone else would find it: file name, record number, page or paragraph. Items with no source are flagged.
6. Reliability:
   • Verified: confirmed by a record, test or physical evidence.
   • Corroborated: supported by at least one independent source.
   • Single source: one account, nothing confirming it yet.
   • Disputed: contradicted by another source.
7. PEEPO tag: People, Environment, Equipment, Procedures or Organisation. This tells you where your evidence is concentrated.

The counters above the table show the spread by reliability and by PEEPO category. If almost everything is tagged People and Single source, you have interviewed the people closest to the event and stopped there.

Sequence first, causes later.

1. Click Add timeline entry. A new entry copies the time of the previous one, so edit it straight away.
2. When: type the date and time. 14/05/2026 2:05pm, 14 May 2026 14:05 and 2026-05-14 14:05 all work and convert to the same format. Add seconds (14:05:30) where they matter. For older events a date, or just a year and month (2014-11), is enough. Dates are read day first, Australian style.
3. Type: choose the entry type (below).
4. Describe what happened or what condition existed.
5. Use the + evidence dropdown to link the evidence that supports the entry. Remove a link with the × on its tag.

Entries sort by time automatically. Where two neighbouring entries both have times and fall within 24 hours, the gap between them is shown (for example +53 s). Those gaps matter. They show how long there was to intervene.

1. Write the problem statement. Specific: what happened, to whom, when. “Worker injured” is not a problem statement. “Maintenance fitter’s hand drawn into the tail pulley of conveyor CV3 during cleaning, 14 May 2026” is.
2. Name the chain after the line of causation it follows, for example “Why the guard was off”.
3. Answer Why 1: why did the problem occur? Link evidence.
4. Click Ask why again and answer why that was so. Repeat.
5. Click Add causal chain for each separate line of causation.

When to stop

Stop when the answer is a condition the organisation controls. Do not stop when you reach a person. Five is a guide, not a rule. A chain shorter than five gets a tip, and one that ends in blame language gets a red note.

Flags on each answer

• Blame language. Triggered by words such as failed to follow, did not follow, careless, complacent, human error, operator error, inattentive, negligent, ignored, violated, should have known, reckless. The question to ask instead: what made that action likely, easy or reasonable at the time?
• Unsupported. An answer with no linked evidence. Link something or treat it as a hypothesis to test.

At Level 2 and above, a single chain gets a prompt. Serious incidents almost never have one cause.

The bowtie shows what should have prevented the loss of control (left side) and what should have limited the harm once it happened (right side).

1. Hazard: the energy or condition with potential for harm. For example “Rotating tail pulley on a powered conveyor”.
2. Top event: the moment control was lost. Not the injury. For example “Person’s body part enters the nip point while the conveyor is energised”.
3. Click Add threat for each thing that could release the hazard.
4. Under each threat, click Add preventive barrier. Barriers are numbered P1, P2 and so on.
5. Click Add consequence for each outcome that could follow the top event, and Add mitigating barrier under each. These are numbered M1, M2 and so on.
6. For every barrier set the control type (Elimination, Substitution, Engineering, Administrative, PPE), the status, and link evidence.

Rate each barrier by what the evidence shows it did on the day, not by what the procedure says it does. A lockout procedure nobody could practically follow is Failed or Degraded, not Effective.

The summary strip counts barriers mapped, failed or absent, degraded, and the percentage at engineering level or higher. In the diagram, rounded barriers are administrative or PPE. If most of your barriers are rounded, the system was relying on people being perfect under pressure. Any threat or consequence protected only by administrative or PPE barriers is flagged.

PEEPO forces the analysis beyond the people and equipment nearest the event. Add factors under each of the five headings with Add factor.

Classify each factor as Immediate cause, Contributing factor, Root cause or Not a factor, and link evidence.

• An empty column is a finding. Either you ruled it out with evidence (add a factor marked Not a factor saying why) or you have not looked.
• An empty Organisation column is a red note. That is where root causes usually sit.
• A People factor classed as a root cause in blame language is a red note. Trace it back to the system that shaped the behaviour.

Root causes

A root cause is a system condition that, if fixed, stops this class of incident, not just this one event.

1. Look at Candidates pulled from your analysis. The tool collects the last answer in each why chain, every Failed or Absent barrier (Level 2+) and every PEEPO factor classed as Root cause (Level 3).
2. Click Promote on a candidate to turn it into a root cause, or Add root cause to write one from scratch. Usually you will combine several candidates into one clearer statement.
3. Link evidence.
4. Tick each of the four tests only if it is true:
   • Fixing it would prevent recurrence, not just this exact event.
   • It sits within the organisation’s control to change.
   • It is supported by linked, reliable evidence.
   • It describes a system condition, not a person’s act or attribute.

When you promote a candidate, the tool pre-ticks the evidence test if evidence is linked and the system test if no blame language is found. The other two are yours to judge. Each root cause shows Passes X/4 tests. Anything under 4 needs rewriting or more evidence.

Corrective actions

1. Click Add action and describe what will change.
2. Set the Control level: Elimination, Substitution, Engineering, Administrative or PPE.
3. Name an Owner (a role is fine) and a Due date or trigger, such as “Before restart”.
4. Use + root cause to link the action to the root cause(s) it addresses. Every root cause needs at least one action.

Push every action as high up the hierarchy of controls as you can. The tool flags any action that relies on words like retrain, remind, toolbox, awareness, be careful, reinforce, counsel, discipline, refresher, re-induct, signage, whatever control level you give it.

The report page shows the full investigation report exactly as it will download: case details, summary, root causes, corrective actions, timeline, causal chains, bowtie barrier tables (Level 2+), PEEPO table (Level 3) and the evidence register. Evidence IDs appear in square brackets after each finding.

1. If a red DRAFT box appears at the top, the investigation check still has critical notes. They print in the report too, so nobody mistakes a draft for a finished investigation. Fix them before you issue it.
2. Click Download Word (.docx) for an editable A4 report with page numbers. Add your organisation’s logo, sign-off block or distribution list in Word.
3. Click Print or save as PDF, then choose Save as PDF as the printer. The tool’s menus and panels are left off the page.
4. Click Export case (.json) to keep the editable case itself. The Word and PDF files cannot be imported back into the tool.

If a download does nothing inside this page, open the Workbench full screen and try again. Some browsers and workplace settings block downloads from embedded windows.