Metal roof screw pull-out strength is set mainly by the substrate the threads bite into, not by the screw you buy. Pull-out (withdrawal) is the load that strips the threads out of the purlin, rafter, or deck. Every published equation for it, in steel and in wood, multiplies substrate thickness or substrate density by fastener diameter. Change the substrate and the number moves several hundred percent. Change the screw and it moves a few dozen.
This page is the engineering side of the question: how withdrawal capacity is calculated, which failure mode governs on your assembly, and how an uplift pressure becomes a screw spacing pattern. If you want ranked product recommendations instead, see our companion buyer guide to the best metal roofing screws.
By The Roofing Brief Team. Last reviewed: August 2026.
What pull-out strength measures on a metal roof
Pull-out strength is the tensile load, in pounds-force, that pulls a screw’s threads out of the member they are threaded into. On a metal roof it is the fastener’s share of wind uplift. It is a property of the connection, not of the screw alone: the same screw has a different pull-out value in a 12 gauge steel purlin, a spruce 2×4, and a 7/16 inch OSB deck.
Two things follow from that. First, a screw box printed with a single pull-out number is describing one tested condition, not the screw. Second, you cannot compare two screws honestly unless both numbers came from the same substrate, the same substrate thickness or species, and the same embedment depth.
Pull-out, pull-over, and screw tension: three ways the connection fails
An exposed fastener metal roof in uplift can fail three ways, and design uses the smallest of the three. Pull-out strips the threads from the substrate. Pull-over tears the panel over the head and washer, leaving the screw in place. Screw tension snaps the shank at the root diameter. ICC-ES evaluation reports for tapping screws publish all three, and instruct designers to take the least of them.
| Failure mode | What physically happens | Governed by | Published as |
|---|---|---|---|
| Pull-out (withdrawal) | Threads strip out of the purlin, rafter, or deck | Substrate thickness or density, thread diameter, embedment | Pnot in steel (AISI S100 J4.4.1); W in wood (NDS Eq. 12.2-2) |
| Pull-over (pull-through) | The roof panel tears over the screw head and washer | Panel thickness, effective washer diameter, panel tensile strength | Pnov in steel (AISI S100 J4.4.2); WH in wood (NDS Eq. 12.2-6) |
| Screw tension | The shank breaks at the root diameter | Screw material and root diameter | Pnts, reported by the manufacturer or by independent lab test |
Which one governs is assembly specific. On a heavy panel over a thin purlin, pull-out usually wins the race to failure. On a 29 gauge panel over a thick purlin, the panel tears over the washer first. That difference is also part of why standing seam and exposed fastener metal roofs behave differently in wind: a concealed clip system does not put the withdrawal load through a hole in the panel face at all.
How the substrate governs pull-out capacity
Every code-recognized withdrawal equation is dominated by a substrate term. In cold-formed steel it is the thickness of the member the screw threads into. In sawn lumber it is the specific gravity of the species, squared. Fastener diameter appears in both, but only to the first power, which is why doubling your substrate does far more than upsizing your screw.
Light-gauge steel purlins
For steel-to-steel screw connections, AISI S100-16 (2020) with Supplement 2 gives the nominal pull-out strength of the sheet per screw in Section J4.4.1 as Pnot = 0.85 tc d Fu2, where tc is the lesser of the depth of penetration and the thickness of the member not in contact with the screw head, d is the nominal screw diameter, and Fu2 is that member’s tensile strength. The provisions apply to screws from 0.08 to 0.25 inch diameter, with an allowable strength design safety factor of 3.00.
Read that equation carefully: capacity is linear in purlin thickness and linear in screw diameter, and there is no term for how good the screw is. Two screws of the same diameter, one premium and one commodity, calculate identically. Only a tested value from an evaluation report can distinguish them, and pull-out is calculated from AISI S100 in most reports anyway.
ICC-ES report ESR-1976 for ITW Buildex TEKS self-drilling fasteners, reissued July 2025, tabulates allowable pull-out (Pnot divided by a safety factor of 3.0) for 45 ksi steel. A #12 screw (0.216 inch nominal diameter) is listed at 99 pounds-force into a 0.036 inch member and 689 pounds-force into a 0.250 inch member. The report also notes that for 58 ksi steel you multiply by 1.29, and for 65 ksi steel by 1.44.
Solid wood purlins and rafters
The 2018 National Design Specification for Wood Construction gives the wood screw reference withdrawal design value in Equation 12.2-2 as W = 2850 G² D, in pounds per inch of thread penetration into side grain, where G is the species specific gravity and D is the screw diameter. Capacity scales with the square of density, so species selection matters more here than anywhere else in the assembly.
Using NDS Table 12.3.3 specific gravities, a #12 screw (D = 0.216 inch) has a reference withdrawal value of 186 pounds per inch of thread penetration in Southern Pine (G = 0.55), 154 in Douglas Fir-Larch (G = 0.50), 114 in Hem-Fir (G = 0.43), and 108 in Spruce-Pine-Fir (G = 0.42). Same screw, same hole, 72 percent spread from the framing species alone.
One NDS rule is routinely violated on job sites: Section 12.2.2.3 states that wood screws shall not be loaded in withdrawal from end grain, with an end grain factor of zero. A screw driven into the cut end of a rafter or into the edge of a board has no calculable withdrawal capacity at all.
Plywood and OSB decks
Here is the part most pages skip. The NDS withdrawal equations are written for penetration into the side grain of a wood member. The NDS does not publish a reference withdrawal design value for screws in plywood or oriented strand board. Panels appear in Table 12.3.3B only for dowel bearing strength, where plywood of unknown species is assigned G = 0.42 and OSB is assigned G = 0.50.
That absence is the honest answer to “what is the pull-out strength of a metal roofing screw into 7/16 inch OSB.” There is no code table for it. Two practical consequences follow. Embedment is capped by panel thickness, so a screw in a 7/16 inch deck can never develop the thread engagement it would get in a 2x purlin. And the residential code responds by testing the assembly rather than calculating the fastener: IRC Section R905.10.5 requires metal roof panels applied to a solid or closely fitted deck to be tested for wind resistance in accordance with FM 4474, UL 580, or UL 1897.
Deck substance is therefore a wind decision, not only a cost decision. Our comparison of OSB vs plywood roof decking covers the thickness and moisture behavior that sit underneath this.
How much does the screw itself actually change the number?
We ran the published ESR-1976 allowable pull-out table two ways to isolate the variables. Holding the substrate at 0.036 inch steel and upsizing the screw from #10 to 1/4 inch raises allowable pull-out from 87 to 115 pounds-force, a 32 percent gain. Holding the screw at #12 and thickening the purlin from 0.036 inch to 0.250 inch raises it from 99 to 689 pounds-force, a 596 percent gain.
| Variable held constant | Variable changed | Allowable pull-out, per ESR-1976 (45 ksi steel, factor of safety 3.0) | Change |
|---|---|---|---|
| Substrate: 0.036 in steel | Screw #10 to 1/4 in | 87 to 115 lbf | +32% |
| Screw: #12 (0.216 in) | Substrate 0.036 to 0.105 in | 99 to 289 lbf | +192% |
| Screw: #12 (0.216 in) | Substrate 0.036 to 0.250 in | 99 to 689 lbf | +596% |
The same table exposes which mode governs. A #12 hex washer head screw through a 0.036 inch member has an allowable pull-over of 336 pounds-force but an allowable pull-out of 99 pounds-force into a 0.036 inch purlin. Pull-out governs by more than three to one. Flip the assembly to a thin residential panel over a thick purlin and pull-over becomes the limit, because pull-over scales with the panel thickness under the head.
Thread type matched to substrate
Thread geometry does not create withdrawal capacity, it lets the screw reach and hold the capacity the substrate already has. The wrong thread either fails to cut cleanly, which strips the hole, or removes too much material, which shrinks the effective embedment.
- Wood purlins and rafters: a sharp type 17 point with coarse, deep, widely spaced threads. Coarse threads displace more wood fiber per turn and resist backing out in low density species.
- Light-gauge steel purlins: a self-drilling point with fine, machine-style threads. The drill flute must clear the hole before the threads arrive, or the screw work-hardens the hole and strips it.
- Drill point sizing: point numbers are rated by total steel thickness the screw can drill. Undersizing the point on thick steel is the most common cause of a stripped connection that tests fine on paper. Our reference on Teks self-drilling roofing screws lists the point numbers against their drilling capacities.
- Plywood and OSB decks: coarse thread, and pick a length that keeps the point inside the panel or lands in framing. A screw that pierces a 7/16 inch deck and dangles has almost no engagement.
Embedment depth: only engaged thread counts
Withdrawal capacity is proportional to engaged thread length, and both governing standards say so explicitly. The NDS multiplies its per-inch value by the thread penetration depth pt, excluding the tapered tip. AISI S100 caps its tc term at the lesser of penetration depth and member thickness, so a screw that stops short of full purlin thickness is designed on the shorter number.
The practical rule that follows: screw length is chosen so that after the panel, any closure, and the washer, the remaining thread reaches the target embedment. Manufacturers of metal-to-wood roofing screws commonly publish their tested values at a stated penetration into the wood member, so a value quoted at one inch of penetration is not valid at half an inch. Check the footnote on the table, not the headline number.
Overdriving: the seal fails long before the withdrawal capacity does
Overdriving and pull-out are two different problems and conflating them causes bad diagnoses. Overdriving crushes the EPDM washer and dishes the panel around the head. That destroys the water seal immediately and reduces the effective pull-over diameter, because a dished panel no longer bears flat under the washer. Withdrawal capacity, which lives in the threads down in the substrate, is largely untouched until the driver actually strips the hole.
So an overdriven roof usually leaks years before it blows off, and an underdriven roof holds water out until the gasket is not compressed enough to seal. AISI S100 Section J4.4 sets minimum head or washer dimensions for screws carrying tension: a washer diameter of at least 5/16 inch, with nominal washer thickness of at least 0.050 inch when the panel is thicker than 0.027 inch. Those minimums assume the washer is intact and seated flat, which an overdriven screw is not.
The diagnostic question on a failed roof is therefore which mode you are looking at. Rings of rust and dished panels around heads point to overdriving and pull-over. Screws standing proud with clean threads, or missing entirely with intact panel holes, point to withdrawal from the substrate.
Turning an uplift load into a fastener spacing pattern
Fastener spacing is the output of a load calculation, not a habit. The chain runs from wind pressure to tributary area per fastener to spacing, and it is checked separately in each roof zone. Design of the panel system itself is done by test: IRC Section R905.10.5 requires structural standing seam systems to be tested to ASTM E1592 or FM 4474, and structural through-fastened systems to ASTM E1592, FM 4474, or UL 580.
- Get the design uplift pressure for components and cladding from ASCE 7 Chapter 30 (or the prescriptive tables the local code adopts), for the building’s exposure category, mean roof height, and effective wind area.
- Split the roof into zones. ASCE 7-22 uses zones 1, 2, and 3, with pressure rising from field to edge to corner. The width of the edge and corner strips is set by the edge strip dimension “a”, commonly published as 10 percent of the least horizontal building dimension with a 3 foot minimum. Confirm the exact definition against the adopted edition, since it has changed between editions.
- Take the governing allowable fastener capacity: the least of allowable pull-out, allowable pull-over, and allowable screw tension, from the evaluation report for your screw at your actual substrate and embedment.
- Divide capacity by pressure to get the maximum tributary area each fastener may carry, in square feet.
- Resolve that area into a pattern. Purlin or rafter spacing fixes one dimension, so the tributary area sets the number of fasteners across the panel width, which is the rib count you screw at.
- Tighten in zones 2 and 3. Because corner pressure exceeds field pressure, the same fastener needs a smaller tributary area there, which usually means screwing every rib at edges and corners where the field is every other rib.
- Check the geometry limits. AISI S100 requires fastener spacing of at least 3d and edge and end distance of at least 1.5d for screws to be considered fully effective.
For how the resulting pattern is actually laid out on the panel, including the flat versus rib question and eave, ridge, and lap fastening, see our screw pattern for metal roofing guide.
How to read a published pull-out value without getting burned
A withdrawal number without its test condition is not usable. Anyone quoting “these screws hold 542 pounds” is quoting one row of one table. Before you carry a number into a calculation, confirm five things.
- Substrate and thickness or species. Steel values are tied to a design thickness in inches and a tensile strength in ksi. Wood values are tied to a specific gravity.
- Embedment. The penetration depth the value was established at, since capacity is per inch of engaged thread.
- Allowable or nominal. Allowable strength design values already carry a safety factor (3.0 for AISI S100 screw connections). Nominal values do not.
- Calculated or tested. Many report tables are calculated from AISI S100 rather than physically pulled, which is legitimate but means the number is a property of the geometry.
- Source document. An ICC-ES or IAPMO evaluation report number with an issue date beats a marketing page. ESR-1976, for example, was reissued July 2025 and is subject to renewal July 2027.
One comparison trap deserves its own warning. NDS wood values and AISI allowable values are not on the same basis and should not be placed side by side without adjustment. NDS reference design values are adjusted by factors including load duration, which for wind is a significant increase, while AISI allowable values already embed a fixed safety factor. Converting between them is the engineer of record’s job, not a spreadsheet shortcut.
Frequently asked questions
What is the pull-out strength of a metal roofing screw?
There is no single value. Pull-out strength is a property of the connection, so it depends on the substrate, its thickness or specific gravity, the screw diameter, and the embedment depth. Published allowable pull-out for a #12 self-drilling screw in 45 ksi steel ranges from 99 pounds-force at 0.036 inch thickness to 689 pounds-force at 0.250 inch, per ICC-ES report ESR-1976 (2025). Use the report for your specific screw.
Does a bigger screw give more pull-out strength?
Only modestly. In both governing equations, capacity is linear in screw diameter but linear in substrate thickness (steel) or proportional to the square of specific gravity (wood). Using ESR-1976 values, upsizing from #10 to 1/4 inch in a 0.036 inch member raises allowable pull-out by about 32 percent, while thickening the purlin from 0.036 to 0.250 inch raises it by roughly 596 percent with the screw unchanged.
What is the difference between pull-out and pull-over?
Pull-out strips the threads out of the substrate, leaving the panel and screw head intact and the screw travelling upward with the panel. Pull-over tears the panel over the head and washer, leaving the screw threaded in place. AISI S100-16 calculates them separately, as Pnot in Section J4.4.1 and Pnov in Section J4.4.2, and design uses whichever is smaller along with the screw’s own tension capacity.
How deep does a metal roofing screw need to go into a purlin?
Deep enough to develop the embedment the published value was established at. The NDS multiplies its per-inch withdrawal value by thread penetration excluding the tapered tip, and AISI S100 caps its tc term at the lesser of penetration depth and member thickness. Manufacturer tables state a required penetration, and a value quoted at one inch into wood is not valid at a shallower depth.
Does overdriving a screw reduce pull-out strength?
Not directly. Overdriving crushes the EPDM washer and dishes the panel, which destroys the water seal and reduces effective pull-over resistance because the washer no longer bears flat. The threads in the substrate are largely unaffected until the driver actually strips the hole. In practice an overdriven roof leaks long before it releases, so overdriving is primarily a leak defect and secondarily a pull-over defect.
What is the pull-out strength of a screw in OSB or plywood?
The National Design Specification does not publish reference withdrawal design values for screws in plywood or oriented strand board; its withdrawal equations apply to side grain of a wood member. Panels appear in NDS Table 12.3.3B only for dowel bearing. That is why IRC Section R905.10.5 requires metal panels over a solid or closely fitted deck to be tested as an assembly to FM 4474, UL 580, or UL 1897.
Which test standards apply to metal roof uplift?
Per IRC Section R905.10.5, structural standing seam metal panel roof systems are tested to ASTM E1592 or FM 4474, and structural through-fastened systems to ASTM E1592, FM 4474, or UL 580. Metal panels over a solid or closely fitted deck are tested to FM 4474, UL 580, or UL 1897. Individual screw connection strengths in cold-formed steel come from AISI S100 Section J4 or a product evaluation report.
Can I calculate screw spacing from an uplift pressure myself?
You can size a trial pattern by dividing the governing allowable fastener capacity by the design uplift pressure to get a maximum tributary area per fastener, then tightening that area in the edge and corner zones. In many jurisdictions the final pattern still needs an engineer of record, because zone dimensions, exposure category, and load duration adjustments vary by adopted code edition and building geometry.