How to Run a Daily Excavation Inspection That Would Hold Up

Why this matters

29 CFR 1926.651(k)(1) requires a competent person to inspect the excavation, the adjacent areas and the protective systems prior to the start of work and as needed throughout the shift, and after every rainstorm or other hazard-increasing occurrence. It does not require you to write any of it down. That absence is the useful fact in this article, because it tells you what a record is for. It is not there to satisfy the clause. It is there to fix a decision to a time, and to make change visible from one morning to the next, which is the only thing that separates an inspection from a look.

Most inspection sheets in circulation are condition checklists: is the ladder there, is the box in, any cracks. Those fields all read "yes, yes, no" on the morning before a failure. The fields that would have caught it are about what happened since the last entry, and almost no sheet has them. Nobody enters the excavation until this is finished, including the person doing it.

The record, entry by entry

Ten entries, in the order the walk actually goes: know what changed before you look, then classify and measure, then check the system against both, then read the faces, then read everything around them, then the way out, then the air, then decide.

1. Who, when, and which trigger. Name, designation, clock time to the minute, and which of the three triggers this entry is - pre-shift, during-shift, or event-driven. A sheet with one line per day cannot show that the rain inspection happened, and the rain inspection is the one that matters most.

2. What has happened since the last entry. Precipitation with an amount and a window, freeze-thaw, any vibration source (a compactor, a roller, traffic on plates, pile driving, blasting), any equipment that worked at the edge, any spoil added or moved, any adjacent excavation opened. These are not colour. Appendix A to Subpart P disqualifies soil from Type A for vibration and for previous disturbance, and a surcharge added at the edge takes the protective system outside the tables it was selected from. A blank here is the single most expensive blank on the sheet.

3. Today's soil class, with the two analyses that produced it. Visual finding and manual method and result, per Appendix A. Never "same as yesterday". Where the material changes along the run, the class changes along the run, and the sheet gets two entries with stations.

4. Depth at the deepest station. Measured from the highest adjacent grade, with the station noted. The tallest wall governs, and on a crowned street or a cross-falling corridor it is not the one you are standing next to.

5. Protective system, by identity, checked against 3 and 4. For a shield, the box and its tabulated depth at today's class. For shoring, member size and spacing against the appendix row or the tabulated data. For a slope, the measured layback against the class. If 3 or 4 moved, this line is re-derived, not copied.

6. Faces and edges. Tension cracks with distance back from the edge and length, spalling, slough at the toe, bulging, water on the face, undermining. Read from grade, standing outside the influence zone.

7. Spoil and surface loads. Toe distance, pile height, and anything new inside the influence wedge - a pipe stack, a plate, a parked truck, a set of outriggers.

8. Access and egress. Ladder stations, and the lateral travel from the furthest work point to the nearest means of egress along the path a person can actually walk, which is not the same as the distance along the trench when a shield end plate is in the way.

9. Atmosphere trigger review, and results if tested. Whether the conditions in 1926.651(g)(1)(i) exist, recorded either way. "Reviewed, trigger not met, reason" is a valid and useful entry.

10. The decision, signed. One of three: work proceeds, work proceeds after these named corrections, or nobody enters. A record whose decision line has never once said the third thing on a job that runs open cuts is telling you something about the record, not about the ground.

One morning, filled in

Job: 120 ft of open trench, 3 ft wide, residential street, shielded, sanitary lateral.

1. Who, when, trigger. 06:42, pre-shift, competent person designated in writing for this project. Prior entries: Tuesday 06:35 and Tuesday 14:10.

2. Since last entry. Rain 22:00 to 01:00, 0.35 in on the site gauge. Plate compactor operated on the parallel backfill lane, 18:00 to 19:30 Tuesday, per the rental ticket. Two spoil loads added at the north end Tuesday afternoon.

3. Class. Visual: faces damp to about 3 ft, no free water, no seeping, no fissures. Material previously undisturbed except at the north end, where an old water service trench crosses. Manual: thumb penetration on a sample taken from the machine's last bucket at grade, indented with effort. Class B, stations 0+00 to 0+80. Class C, stations 0+80 to 1+20, previously disturbed material at the old crossing.

4. Depth. 9.4 ft at station 1+05, measured from the crown side, which sits 0.5 ft above the gutter side.

5. System. Two boxes, 8.0 ft tall. Tabulated depth for Type C confirmed at or above 9.4 ft on the data sheet for these boxes. Box height 8.0 ft against 9.4 ft of cut leaves a 1.4 ft face above the box, and the boxes as set stand LEVEL with the top of that vertical face rather than the 18 in above it Appendix B requires. Benching is not an option through the north section either, because that section classified Type C and Appendix B does not permit benching in Type C at all. To close both at once: slope the upper portion at 1.5 to 1 over at least 2.9 ft of depth, which drops the vertical portion to 6.5 ft, leaves the 8.0 ft box standing the required 18 in above it, and consumes at least 4.4 ft of run each side. Correction required.

6. Faces. No tension cracks anywhere on the run. Slough at the toe, north end, roughly 6 in deep over about 8 ft of length, in the disturbed material.

7. Spoil. Toe at 6 ft from the edge at the north end. It was at 9 ft Monday; Tuesday's two loads moved it. Screening wedge for Type C at 9.4 ft, projected back from the bottom edge, is 9.4 x 1.5 = 14.1 ft. The toe is 8.1 ft inside the wedge. Correction required.

8. Egress. Ladders at 0+25, 0+75 and 1+15. Along the open trench that passes: furthest points travel 25 ft, 25 ft, 20 ft and 5 ft to the nearest ladder, all at or under the 25 ft in 1926.651(c)(2). But the box end plate at 0+80 blocks passage, so 0+80 to 1+20 is a 40-ft compartment with one ladder at 1+15, and a worker at 0+80 travels 35 ft. Fails. One ladder at 1+00 fixes it: 20 ft from 0+80 and 20 ft from 1+20.

9. Atmosphere. Trigger reviewed. No landfill, no known contaminated ground, the sanitary line is not open, the pump is electric off the service and no engine is running at the edge. Trigger not met, not tested.

10. Decision. Nobody enters. Four corrections, then re-inspect:

• Re-cut the upper portion of the north section as a 1.5 to 1 slope over at least 2.9 ft of depth, dropping the vertical portion to 6.5 ft so the 8.0 ft boxes stand the required 18 in above the top of the vertical side. No benching in Type C. Machine working from grade, nobody in the cut.
• Pull the north-end spoil toe back beyond 14.1 ft, machine working from grade, nobody in or beside the cut, with a stop log and a spotter at the edge per 1926.651(f) because the operator has no direct view of the edge over the pile.
• Set a ladder at station 1+00, lowered from grade into the box, secured, extending 3 ft above the landing surface, before any entry.
• Re-read the north-end toe slough from grade after the spoil is pulled, because unloading the edge can free material that was being held in place.

07:30 re-inspection. North-section upper portion re-cut to a 1.5 to 1 slope over 3.0 ft of depth, vertical portion measured at 6.4 ft, box tops standing 1.6 ft above the top of the vertical side, run measured at 4.6 ft each side. Toe pulled to 15 ft. Ladder set at 1+00 and secured. Slough re-read from grade: no growth, no new material at the toe, no crack behind the edge. Work released 07:38. Total delay 56 minutes.

What that morning actually turned on. Tuesday's two entries both read clean, and both were correct when they were written. Every condition found Wednesday was created after Tuesday's last entry, and every one of them was already documented somewhere else in the company - a rain gauge, a rental ticket, a delivery ticket. Entry 2 is the field that pulls those three records onto the sheet where the decision gets made. Without it, the inspector arrives, sees a tidy trench, and writes "OK", and both findings are invisible from grade until one of them stops being invisible.

The failure mode. Note that neither finding was a crack. The classification split at the old service crossing came from a records question, and the egress failure came from geometry the sheet's tick-box version ("ladder present: yes") could never surface. A checklist built around conditions catches the ones that have already announced themselves.

Auditing your own records once a month

The inspection is not the artifact; the run of inspections is. Pull a month of sheets for one job and look for six tells, none of which require you to know anything about the ground:

• Identical clock times. 07:00 every day means the time is being written, not read.
• The same class through a rain event. Water freely seeping puts material in Type C by Appendix A's own definition, so a class that never moves across a wet week is a class that is being copied.
• No split classifications, ever. Real ground changes along a run. A job that produced one class for 400 feet either had remarkably uniform material or one entry that got reused.
• No entry that ever produced a correction. Over a month of open cut, zero corrections is not a good record, it is an incomplete one.
• Signatures on days the person was not on site. Cross-check against time records once; you only need to do it once.
• No during-shift or event-driven entries at all. 1926.651(k)(1) names three triggers and most sheets only ever show one.

Two things this audit is for. The first is operational: a record that cannot show change cannot support the re-decision the inspection exists to make, and you would rather find that in an office than on a wall. The second is evidentiary, and it should be stated plainly rather than promised - a contemporaneous record signed at a stated time by a designated person is a dated position on what was known and decided that morning, which is worth having. What legal weight it carries in any particular dispute is a question for your own attorney, not something an article can settle. The operational value stands on its own.

References

• 29 CFR 1926.651(k), inspections by a competent person prior to the start of work, as needed throughout the shift, and after every rainstorm or other hazard-increasing occurrence, and removal of exposed employees at (k)(2)
• 29 CFR 1926.651(c)(2), means of egress in trench excavations 4 feet or more deep, located to require no more than 25 feet of lateral travel
• 29 CFR 1926.651(f), warning systems for mobile equipment operating adjacent to an excavation; 1926.651(g)(1)(i), atmospheric testing triggers
• 29 CFR 1926 Subpart P Appendix A, soil classification by visual and manual analysis, including the Type A exclusions for vibration and previous disturbance
• See related: universal-what-a-competent-person-is-required-to-be-able-to-do; What Access and Egress Has to Be in Place Before Anyone Enters