What an Anchorage Has to Be Rated For and Who Decides
Why this matters
If you cannot name which acceptance route an anchorage is on and who signed for it, do not clip to it, and do not go to the edge to look at it. Evaluate a candidate anchorage from a protected position or before any exposure exists: from inside a guardrail, from a hatch, from the ground with binoculars, or from the drawings.
An anchorage is the only part of a fall arrest system that you usually did not buy and cannot read a rating off. The harness has a label. The lanyard has a label. The thing at the top of the line is a piece of somebody's building, and the question "is that strong enough" has no answer that lives in the object. It only has an answer that lives in a route and a person.
First, the scoping clause, because this one gets people killed
A fall-protection anchorage is not the same subject as a concrete fastener anchor. This library has several strong cards on selecting and installing anchors for hanging equipment. Those rate a fastener for a sustained service load in a base material, under a design basis that already carries its own factor and that assumes a static, predictable, mostly downward load applied through a known bracket.
A fall-protection anchorage is rated for a single dynamic event: a shock load, arriving fast, often with a large lateral component, sometimes with the line running over an edge that changes its direction. A fixture anchor's published allowable load is not a fall-arrest rating and cannot be converted into one by arithmetic, by anyone except the engineer who designs that specific connection. If a reader takes one sentence off this card, take that one, because tying a lifeline to a threaded rod that was sized to hang a fan coil is a real thing that real techs do.
What the number is rating, and per what
For construction, 29 CFR 1926.502(d)(15) requires that anchorages used for attachment of personal fall arrest equipment be independent of any anchorage being used to support or suspend platforms, and be capable of supporting at least 5,000 pounds per employee attached, or else be designed, installed and used as part of a complete personal fall arrest system that maintains a safety factor of at least two, under the supervision of a qualified person. General industry carries closely matching anchorage criteria at 29 CFR 1910.140(c)(13), including the same two acceptance routes.
Three things in that sentence are routinely misread.
Per employee attached. Two techs on one anchorage is not one anchorage question, it is two. If the answer for one is "just about," the answer for two is no.
Capable of supporting. Not "weighs 5,000 pounds," not "held a truck once." Capacity in the direction the load will arrive, through the whole load path, including the fasteners and the member the fasteners land in.
Independent of any platform anchorage. The point on the parapet that holds your suspended scaffold is disqualified from also holding your lifeline, by name, in the standard.
Positioning device systems are rated differently and it is worth knowing so you do not mix them up: 29 CFR 1926.502(e) sets anchorages for positioning devices at twice the potential impact load or 3,000 pounds, whichever is greater. Different system, different duty, different number.
The two routes are alternatives, not a ladder
There is a persistent belief that 5,000 pounds is the requirement and that engineering is what you do when you cannot meet it. That is backwards in a way that matters.
Route one, prescriptive. Show 5,000 pounds per attached worker. This route exists so that a crew can tie off to an obviously massive structural member without paying an engineer to look at every one. The number is deliberately blunt and deliberately large, because it is standing in for an analysis nobody is going to do at seven in the morning.
Route two, engineered. A qualified person designs the anchorage as part of the complete system to a safety factor of at least two on the forces that system will actually produce. This route can land at a smaller number than 5,000 pounds and be entirely correct, because the analysis is real where the 5,000 was a substitute for one.
The routes do not stack and you do not get to average them. You are on one or the other, and the practical consequence is that "it is probably close to 5,000" is not a position. Either you have a documented basis for the prescriptive figure or you have a qualified person's design. Everything in between is a guess wearing a number.
What comes off the list, and the reason for each
This is the section worth the reading time, because most rooftop tie-off decisions are decided by what a tech excludes rather than by what they select.
Vent stacks, plumbing vents and flue pipe. Rated to carry themselves and to stay weathertight. Frequently sheet metal into a roof jack with no structural connection to anything.
Conduit, cable tray and their supports. Sized for the weight of what is inside them plus a code-mandated margin. Neither the tray nor its rod hangers has any basis for a shock load.
Ductwork, the rooftop unit, and the unit curb. The unit's mass is not its holding capacity. The load path from a curb clip runs into sheet metal, then into fasteners, then into a roof deck, and the deck is the term nobody looks at.
Guardrail posts and the top rail. Guardrail systems in construction are rated to withstand a 200-pound force applied at the top rail under 29 CFR 1926.502(b). That is a factor of twenty-five below the prescriptive anchorage figure, and it is rated for a push, not for a jerk from below.
A parapet. A 30-inch masonry parapet is a wall, not a member designed for a lateral point load at its top course. It is also usually the very edge you would be loading it toward.
Anything holding up a platform. Excluded by name in 1926.502(d)(15), see above.
Another worker's anchorage, when the capacity was established per worker. Two people on a single-worker rating is a violation and an overload at the same time.
A concrete fixture anchor installed for equipment. See the scoping clause above.
And one that is not off the list but is not on it either, which is the honest category most roofs fall into: structural steel you can see. A bar joist top chord, a purlin, an open-web beam. These carry real load, so the instinct is to accept them. But a bar joist is designed for distributed gravity load along its chord, not for a concentrated shock at one point on it with a sideways component that can roll it. Steel being present is evidence that a qualified person might say yes. It is not the yes.
Who decides, and where each authority stops
The competent person. 29 CFR 1926.32(f) defines this as one who is capable of identifying existing and predictable hazards in the surroundings or working conditions which are hazardous to employees, and who has authorization to take prompt corrective measures to eliminate them. The authority half is part of the definition, not an add-on. A person who can spot the problem but cannot stop the work is not a competent person, they are a witness.
The qualified person. 29 CFR 1926.32(m) turns on a recognised degree, certificate or professional standing, or on extensive knowledge, training and experience demonstrated by having solved problems of that type. This is who the anchorage standard names for route two and for horizontal lifelines.
The registered professional engineer. Not named in the anchorage provision, but this is who answers "can this building take it," because that question is about the structure rather than about the fall protection system. A qualified person on fall protection is not automatically qualified on the load path through somebody's roof deck.
The rule of thumb that holds these apart: the competent person decides whether today's conditions permit the plan, the qualified person designs the system, and the engineer certifies the structure the system hangs off.
Worked example: what a tight lifeline does to its end anchorages
A crew wants a horizontal lifeline spanning 30.0 feet between two end anchorages, so one tech can traverse a run of work. Every arrest on that line produces a body force capped at 1,800 pounds by 29 CFR 1926.502(d) for a system used with a body harness. The question is what the two ends see.
Treat the loaded line as a straight two-force member, load applied at midspan, both ends anchored at the same height. Tension in each leg is the applied force divided by twice the sine of the sag angle measured from horizontal.
Body force at arrest (the 1926.502(d) ceiling) 1,800 lbf
This is a RE-BASING, not an addition: the
1,800 already contains the absorber's
limiting action. The line tension is that
same event expressed in a different member.
Sag angle 15 deg: 1,800 / (2 x 0.2588) 3,478 lbf
Sag angle 10 deg: 1,800 / (2 x 0.1736) 5,184 lbf
Sag angle 5 deg: 1,800 / (2 x 0.0872) 10,326 lbf
Read the third line against the prescriptive number. A body force of 1,800 pounds, on a lifeline installed tight enough to sag only 5 degrees at peak load, puts more than 10,000 pounds into each end anchorage. That is over twice the 5,000-pound prescriptive figure, produced by a system that is fully compliant at the body.
That is the whole reason 29 CFR 1926.502(d)(8) requires horizontal lifelines to be designed, installed and used under the supervision of a qualified person, as part of a complete system maintaining a safety factor of at least two. The prescriptive route was never written to cover a geometry that multiplies the force.
State what this relationship holds constant, because it travels badly. It assumes the load lands at midspan, both ends at equal height, the line straight between the load and each end, and the sag angle is the angle at peak load, not the sag you eyeballed at installation. It neglects the line's own weight, its elasticity, and every dynamic effect. Real lines stretch, which increases the sag angle and reduces the tension, so this calculation errs high, and erring high on an anchorage force is the direction you want. It is a teaching relationship, not a design. The number that goes on the drawing comes from the qualified person and, where the building structure is in question, from a registered professional engineer.
How to verify you have an anchorage rather than a hope
Three questions, written down, before anyone connects.
- Which route is this on? Name it. Prescriptive at 5,000 pounds per attached worker, or engineered to a safety factor of two under a qualified person. If nobody can name the route, the answer is that you do not have an anchorage yet.
- What is the whole load path, and who checked the weakest part of it? The anchor plate, its fasteners, the member, and the member's connection to the structure. A manufactured roof anchor's ID plate rates the anchor. It says nothing about the deck it is bolted through.
- How many people are on it, and does the rating say per worker? Recount at the end of the day, not at the start, because crews grow.
The failure mode reads as diligence from three feet away: a tech finds a substantial-looking steel member, wraps a rated strap around it, clips a rated connector to a rated lifeline, and every component in the chain has a label except the one holding all of them. Labels on the parts you bought are not a rating for the part you found.
References
- 29 CFR 1926.502(d), personal fall arrest system criteria for construction, including anchorage requirements at 1926.502(d)(15) and horizontal lifelines at 1926.502(d)(8).
- 29 CFR 1926.502(b) for guardrail system strength, and 1926.502(e) for positioning device anchorages.
- 29 CFR 1926.32(f) and 1926.32(m), the definitions of competent person and qualified person.
- 29 CFR 1910.140, personal fall protection system criteria for general industry.
- See related: How to Compute Fall Clearance and Why People Get It Short, What Swing Fall Does That a Vertical Fall Does Not, How an Anchor Transfers Load Into Concrete (a different subject: fasteners for hanging equipment).