Noise or Signal? Telling Common Cause from Special Cause — and Buffering for Both

May 27
5 min read

Issue 6 · May 28, 2026

Every project throws the same problem at you, over and over: something goes sideways, and you have to decide what it means. A steel package shows up a week late. An inspection fails. A slab pour runs long. Underneath all of them is the same question — is this just the normal noise of building, or is something actually wrong? Get that call right and you spend your energy where it counts. Get it wrong and you either chase ghosts or miss a fire.

W. Edwards Deming gave us the cleanest way to think about it. In Out of the Crisis (1982), he splits all variation into two buckets. Common cause is the ordinary scatter baked into a stable system — the hundred small things nobody fully controls: traffic, the weather on any given day, a mill’s queue, a crew that’s a body light. It has no single reason, because it isn’t one thing. It’s the system breathing. Special cause is the opposite — a specific, assignable event you can point to: a mill breakdown, a detail stuck in approval, a sub that walked off. One is background hum. The other is a signal.

Mixing them up is where projects bleed. Treat common cause like special cause — launch an investigation, expedite the PO, chew somebody out — every time a number wobbles inside its normal range, and you’ve spent real effort on nothing, while teaching your team to hide the wobble next time. Treat special cause like common cause — shrug off the one delivery that’s five weeks out as “steel’s always a little late” — and you’ve let a real problem ride until it’s a crisis.

So how do you tell which one you’re looking at? Three questions get you most of the way there. Is it inside the range you’ve seen before? Pull a few months of history — deliveries against promised dates, task durations against plan, whatever it is — and look at the spread. If this point sits inside the band everything else lives in, it’s almost certainly common cause. Can you name a specific reason? Common cause has no single culprit; if you can point to one assignable event — a roll-change, an RFI hold, a failed rig — that’s the fingerprint of special cause. Does it show up everywhere, or just here? A problem that recurs across every job and every supplier is built into your system. One that appears once, on one project, right after one event, is a special cause worth running down.

Here’s the part most teams skip: once you know which one you’ve got, you manage them differently — and the tool for both is buffer. Common cause never goes away, so you don’t react to it, you absorb it. Size a buffer to the normal range and let it do its job. If deliveries routinely swing three days, your sequence needs at least that much float built in, so a normal late load costs you nothing instead of triggering a fire drill. The buffer is what lets you stop reacting to noise — it turns a “problem” back into what it always was: expected.

Special causes need buffer too, but a different kind. You can’t size routine float to a five-week mill outage, so you carry a separate contingency for the hits you can’t predict, and you cut your exposure where you can: release long-lead details early, dual-source the critical packages, stage the steel you can’t afford to wait on. Then you build the habit of catching special causes fast — because the whole point of protecting the system with buffer is that it frees your attention to spot the real signal when it shows up, and respond before it eats the contingency you set aside.

That’s the discipline: a buffer sized to common-cause scatter so you quit firefighting the noise, a contingency held for the special-cause hits you can’t foresee, and the judgment to tell the two apart so you always know which one you’re spending. The crews that run smooth aren’t the ones with no variation. They’re the ones who stopped reacting to the variation that doesn’t matter and built room to handle the variation that does.

Something to consider. Both halves of this — discerning common from special cause, and sizing a buffer to match — come down to reading your own history, and that history is sitting in your delivery logs, daily reports, and schedules, mostly unread. This is where AI helps. Feed it a few months of records across a couple of projects and ask two things: which delays recur everywhere (common cause — the stuff to buffer for) versus which were one-offs tied to a specific event (special cause), and how wide the normal range actually is. That second answer is your buffer size, grounded in real data instead of a gut number. What AI can’t do is decide how much risk you’re willing to carry, or fix the system that’s producing the scatter. That judgment stays yours.

Steel Market Snapshot

Late May 2026 — HRC clears $1,090 as mills run flat-out and spot tons stay scarce

Hot-Rolled Coil $1,090/ton ▲ Up $10/ton (May 18) · 3–5 wk lead	Plate $1,200–1,250/ton ▲ +$60–80/ton mill hikes announced
Wide-Flange Beams $900–1,150/ton mill $1,100–1,400 service center	Rebar +$20–30/ton ▲ May surcharges (Nucor, CMC)

Nucor pushed hot-rolled spot up another $10 on May 18, mill utilization climbed to a fresh multi-year high above 82%, and plate producers are rolling out base-price hikes into strengthening demand. Sheet inventories are the lowest since May 2021.

Hot-rolled coil sits at $1,090/ton after Nucor’s May 18 increase — the latest in a long string of weekly moves higher. SMU’s market average was $1,080/ton FOB mill (east of the Rockies) as of May 12. Lead times are holding at 3–5 weeks, with spot availability described as extremely tight.

Plate continues to firm. Nucor announced base-price increases of at least $60/ton on as-rolled and normalized plate (and $80/ton on quenched-and-tempered), on top of a $10/ton minimum fuel surcharge effective May 1, into extended lead times and strengthening demand.

Wide-flange beams are running roughly $900–1,150/ton at the mill and $1,100–1,400/ton through service centers, after Gerdau and Nucor structural-section increases in late April.

Rebar keeps climbing — Nucor added a $20/ton surcharge on coiled rebar in early May and CMC raised tags $30/ton, on top of April’s $40–60/ton hikes.

What to watch: Mill capability utilization hit 82.2% the week ending May 16 (up from 80.4% on May 2), with year-to-date running 78.3% versus 76.2% a year ago. With the 50% Section 232 rate on full customs value in force since April 6, imports stay priced out and domestic pricing pressure looks set to hold. For precon teams: lock pricing where the schedule allows and plan for lead times to stretch, not soften. For live data, visit our Market Informer page.

Sources & Further Reading

Deming, W. Edwards. Out of the Crisis (reissue). MIT Press, 2018 (orig. 1982). mitpress.mit.edu
Deming, W. Edwards. The New Economics for Industry, Government, Education. MIT Press, 1993. direct.mit.edu
The W. Edwards Deming Institute. “The System of Profound Knowledge (Understanding Variation).” deming.org
Steel Market Update. “Nucor hikes spot HR price to $1,090/ton.” May 18, 2026. steelmarketupdate.com
Steel Market Update. “Plate market participants predict more mill price hikes.” May 8, 2026. steelmarketupdate.com
Steel Market Update. “Plate sources face challenging market forces amidst strengthening demand.” May 21, 2026. steelmarketupdate.com
SteelOrbis. “US raw steel production increases by 1.3 percent — week 19, 2026.” May 2026. steelorbis.com
ScrapMonster. “U.S. Raw Steel Output Up 9.6% YoY; Utilization Hits 80.4%.” May 6, 2026. scrapmonster.com
GMK Center. “Nucor, Gerdau, and Optimus are raising rebar prices in the US.” 2026. gmk.center
White & Case LLP. “United States modifies steel, aluminum, and copper Section 232 tariffs.” 2026. whitecase.com
SteelFlo. “Structural Steel Cost Per Ton in 2026: What Shops Are Paying.” 2026. steelfloai.com

Champion Steel Works — Steel you can trust.