The Most Expensive Minute in the Shop

Three things worth your attention this week.

Productivity upgrade proposals keep quoting heats-per-day gains from bigger transformers, and the arithmetic is usually honest as far as it goes. What the proposals skip: in most shops the variance in tap-to-tap already exceeds the minutes the new transformer would buy, and variance costs nothing to attack. Before pricing megawatts, price a month of honest delay coding. One of these two projects needs a board approval; the other needs a decision.

Every benchmarking table you'll see this year ranks shops by average tap-to-tap, and averages are the wrong lens for it. Two shops at the same 55-minute average — one holding 53-57, one swinging 45-75 — are different businesses: the steady one can promise its caster a sequence, the swinging one promises its caster surprises. Distribution beats average in every conversation that matters downstream.

New Zealand Steel tapped the first heat on its new Glenbrook EAF on August 11 — two years after groundbreaking, holding its 650,000-tonne annual output through the changeover and cutting site coal use by half. First heats make press releases; the next thousand set the shop's tap-to-tap distribution, and the delay-coding habits a new crew builds now are the ones it keeps for a decade.

Tap-to-tap has an anatomy, and most shops have never dissected one honestly. Power-on time is the physics portion: energy required divided by power delivered, and for a given charge and transformer it's remarkably stable — published figures for modern high-power furnaces put it broadly in the 30-40 minute range. Everything wrapped around it — charging, tapping, EBT maintenance, gunning, temperature and sample waits, and the category that decides careers, "unaccounted" — is management. That split matters, because shops that want a shorter tap-to-tap almost always attack the physics, which is expensive, instead of the management, which is nearly free.

Here's the pattern we've watched repeat across shops: pull sixty days of heat logs and reconcile every minute. Power-on tracks the energy model within a minute or two. The scheduled power-off blocks come in close to standard. And then there's a residual — routinely five to eight minutes a heat — that appears in no delay code anywhere. A crane that arrived late but not late enough to log. A ladle not quite ready. A temperature taken twice because the first felt wrong. Thirty seconds of hesitation at each of a dozen transitions. None of it visible, all of it real, and at a heat every hour it compounds into the equivalent of days of lost production a year.

That residual is the most expensive minute in the shop precisely because it's unowned. A logged delay has a name attached and gets fixed; an invisible one gets absorbed into "that's just our tap-to-tap." The fix costs a stopwatch and a level of delay-coding honesty that most delay cultures punish — which is exactly why the minutes are still there. And there's a thermal bill stacked on the productivity bill: minutes with the roof open or the bath idle radiate heat you repurchase at the next power-on, so the invisible minute charges you twice — once in tonnes, once in kWh.

Now the wrong way to find minutes, because it's seductive and we've seen the bill. When pressure comes down to cut tap-to-tap, the minutes that surrender easiest are quality minutes: the flat-bath refining you trimmed, the temperature you estimated instead of measured, the sample you skipped because the sequence was tight, the gunning round pushed to next shift. Each surrender works — the log shows a faster heat — and each sends an invoice somewhere else: a re-blow that costs more than it saved, a chemistry miss caught at the caster, a downgrade, a slag line that ages a campaign early. The heat you saved four minutes on and then reheated is not a faster heat. It's a slower heat with better-looking paperwork.

The counter-case to minute obsession deserves its sentence: tap-to-tap is a meltshop KPI in service of a plant KPI. A furnace sprinting ahead of its caster earns nothing — sequence match sets the tempo that pays, an argument that gets its own issue later this year. Cut the invisible minutes because variance is free money and calm shops run safer. Cut the visible ones only when the whole line can cash them.

One target: find five minutes per shift without touching quality. Five places they hide, checked in this order.

1. The reconciliation gap. Total tap-to-tap minus power-on minus logged delays. Whatever's left is your hunting ground. If it's over four minutes a heat, stop reading and go watch a heat end-to-end.

2. Transitions. Stand at the panel and time the seams: power-off to tap start, tap end to charge start, charge end to power-on. Nobody owns seams, so nobody trims them. Thirty seconds recovered at three seams, every heat, is the cheapest capacity you will ever buy.

3. Waiting on steel data. Minutes between "sample sent" and "result actioned," and between temperature request and reading. If the furnace idles while the lab works, the fix is sequencing the measurement earlier — not skipping it.

4. Crane and ladle readiness. Log one shift of "furnace ready, lift not ready" events. In shops we've seen run this log, it embarrasses someone into a fix within the week.

5. The second measurement. Count repeated temperatures and re-samples. Each repeat is a symptom — probe condition, foam access, operator trust — and every symptom has a cheaper fix than the two minutes it steals per occurrence.

Log the five numbers at handover for two weeks before changing anything. The pattern will tell you which fix pays first — and you'll have the baseline to prove it did.

Next week: the hot heel argument — why the iron you don't tap is earning its keep.

Written by active DRI-EAF operators. Anonymous by necessity, specific by design.