Technical Article · Plant Operations & Cost Management
A field-calibrated framework for tracking where your zinc actually goes, what it costs, and the five operational levers that control it, with no capital spend
This article has been substantially revised following technical review by a galvanizing consultant with more than 30 years of direct plant management and audits of roughly 175 plants over the last four years. Eight substantive corrections were made. Three of them reversed the original conclusion, so if you are working from Revision 1.0, replace it.
Corrected in this revision: (1) zinc consumption by section thickness was inverted; (2) dross and ash were measured against a dependent variable; (3) the temperature-to-dross sensitivity was overstated by roughly an order of magnitude; (4) the flux concentration advice was backwards; (5) the drossing frequency rule ignored plant size; (6) the ASTM A123-24 claim was factually wrong; (7) kettle level was missing from the measurement-error list; (8) by-product recovery values were understated, which had the knock-on effect of overstating the financial case for reducing dross and ash. A full correction log appears at the end.
Added in this revision: a direct assessment of the published percentage benchmarks, and a section on where the zinc money actually sits once the by-product economics are worked through properly. The short answer is not dross and not ash.
Zinc is the single largest variable cost in a hot-dip galvanizing plant. It is also the cost that most plant managers understand least precisely. They know their zinc bill at the end of the month. Few know exactly how much went on the product, how much went into dross, how much went into ash, and whether any of those numbers are normal or abnormal for their work mix.
This article sets out the framework I use across my three plants: the formulas, the field benchmarks, what the numbers mean when they sit outside target, and the operational controls that govern them. It needs nothing more than an accurate scale, a lab iron test kit, a kettle level rule, and a shift log.
One principle runs through all of it. A KPI is only as good as its denominator. Most of the corrections in Revision 2.0 are denominator corrections, not process corrections.
A complete HDG plant KPI system covers four operational domains. Each domain has one primary metric that tells most of the story, and supporting metrics that explain it when it sits outside target.
| Domain | Primary KPI | Supporting Metrics |
|---|---|---|
| Production Volume | Tonnes galvanized per month | Batches per shift · Average batch weight · Days operated |
| Zinc Efficiency | Zinc consumption % (level corrected) | Pickup % · Dross kg/t · Ash kg/t · Net zinc cost per tonne |
| Quality | First-pass rate (%) | Rework batches · Thickness compliance rate · Customer complaints |
| Equipment Effectiveness | Capacity utilisation (%) | Kettle uptime · Unplanned downtime hours · OEE |
None of these require sophisticated data systems. All of them can be calculated from a batch weight log, a zinc stock record, a kettle level log, and a quality inspection register.
Three definitional choices determine whether your zinc numbers mean anything. Settle all three in writing before you calculate a single percentage, and state them on every report you publish. Two plants can run identical processes and report zinc consumption 1.5 percentage points apart purely through these definitions.
Zinc consumption is expressed as a percentage of steel weight, but which steel weight? Black steel is the as-received weight, including mill scale and rust. White steel is the weight after pickling and rinsing, which is lower by roughly 0.5 to 2% on ordinary hot-rolled material and more on heavily scaled or rusted stock. Consumption expressed on a white-steel basis is therefore slightly higher than the same plant expressed on a black-steel basis. The difference is small, but it is systematic, and it is a real source of disagreement when two plants compare figures. Declare the basis on the report header. The field benchmarks in this article are on a white-steel basis unless stated otherwise.
This is the single largest and most commonly ignored distortion in monthly zinc consumption. If the period opens with the bath 12 cm below the reference mark and closes with it 5 cm below, seven centimetres of zinc were added to bath inventory during the period. That zinc was purchased but it was not consumed. Without a level correction it lands in the consumption number as if it had been.
Kettle 13.0 m × 1.5 m internal. Level moves 7 cm closer to the top over the month: 0.07 × 13.0 × 1.5 × 6,620 = 9,036 kg of zinc added to bath inventory, not consumed.
On a plant processing 500 t of steel that month, 9,036 kg is 1.8 percentage points of apparent zinc consumption. A plant reporting 6.3% was actually running at 4.5%. This single error is larger than every process improvement in this article combined, and it works in both directions: run the level down and you can report a flattering number that is not real.
Rule: measure and log kettle level to the nearest centimetre against a fixed reference on the same day and at the same bath temperature every month-end. Apply the correction every period without exception.
Zinc consumption is governed first by surface area per tonne, and surface area per tonne is governed by section thickness. A plant galvanizing 20 mm plate and a plant galvanizing 2 mm tube cannot share a benchmark. This is dealt with in KPI 1.
Every kilogram of zinc that enters the plant is accounted for in one of five places. The first three are the large ones. The last two are small but they are real, and pretending they do not exist is what forces plants to write off an unexplained variance every month.
| Destination | Scale | Notes |
|---|---|---|
| On the product | Dominant | The only destination that earns revenue. Governed by surface area and coating thickness. |
| Bottom dross | Second largest | Fe-Zn intermetallic, mostly zeta phase. Recoverable and sold. |
| Surface ash | Third | Zinc oxide plus flux residue plus entrained metallic zinc. Recoverable at lower value. |
| Bath inventory change | Can be very large in one period | Not a loss. A timing difference. Removed by the level correction above. |
| Unrecovered losses | Small but real | Zinc stripped from jigs, hooks and wire; splash and floor sweepings; hard zinc on kettle walls; zinc on rejected work that is re-dipped. |
Test the old split against measured by-product rates. At 5% consumption, dross at 17.5% of zinc consumed means 8.75 kg of zinc per tonne of steel, which at 95% zinc content is about 9.2 kg of dross per tonne: the very top of the observed field range of 3–9 kg/t, presented as the average plant. Ash at 12.5% of zinc consumed means 6.25 kg of zinc per tonne, which at 85% zinc content is about 7.4 kg of ash per tonne, against a field range of 2–5 kg/t. The 70/20/10 split describes a plant with a real ash problem, not a well-managed one.
Three sets of numbers run through this article and they are not the same kind of number. Treating them as rivals is the mistake. One is a ceiling, one is an aim point, and one is a single operator's year.
| Source | What it is | Denominator | Dross | Ash |
|---|---|---|---|---|
| AGA and GA, commonly cited | Ceiling. Stay under it. | Zinc consumed | ~15% | ~14% |
| F. Priuli, Italian plants | Ceiling, tighter | Zinc consumed | < 13% | < 12% |
| G. Crowley, field figures supplied directly | Aim point. Where a well-run plant sits. | Steel galvanized | 3–9 kg/t | 2–5 kg/t |
| This case study: three plants, 2025 | One operator's measured year | Both, stated separately | 5.60 kg/t · 9.0% | 8.22 kg/t · 11.8% |
The ceiling. Dross at about 15% of zinc consumed and ash at about 14% are the figures most often quoted, and this is the right way to use them: as an upper limit. A plant under those numbers is in acceptable territory. A plant over them has something to find. They describe a population rather than a kettle, which is exactly what makes them serviceable as a limit and poor as an ambition. They are also two independent limits and not a distribution to be added together — a plant sitting at both ceilings at once would be putting only 71% of its zinc on the product, well under the 80 to 90% this article observes.
The aim point. The kilogram-per-tonne figures used throughout this revision are Geoff Crowley's: more than 30 years of direct plant management and audits of roughly 175 plants over the last four years. Zinc consumption 4 to 6% on white steel, dross 3 to 9 kg per tonne of steel, ash 2 to 5 kg per tonne. These were supplied to me directly in correspondence and are not published. They appear here once, as a matter of record. The reviewer of Revision 1.0 and the consultant in this table are the same person, so this is two positions on the denominator question and not four.
The aim point has to be conditioned on the plant, and nobody has reduced that to a formula. A range like 2 to 5 kg per tonne of ash assumes a plant equipped to hit it. One with no drier and no flux heating, running light sheet work, cannot be held to the same figure as one that has both — the conditions that generate ash are installed, not chosen daily. I had assumed this was a gap in my own thinking rather than in the field's. It is not: Crowley reports having made several attempts at a formula accounting for plant configuration — cross-shop versus monorail loop, for instance — and abandoned all of them, and now conditions the target by judgement instead. Two people failing at it independently is not proof that it cannot be done, but it is a reason to stop presenting these ranges as if they applied uniformly. Compare your plant against the range, then adjust for what you actually have installed, and be explicit about that adjustment when you quote the number to anyone else.
The objection to a percentage of zinc is that the denominator is dependent. Ash and dross are measured against zinc consumed, but the zinc inside that ash and dross is part of zinc consumed. Make more ash and both numerator and denominator rise together, so the ratio moves less than the underlying change. Cut average coating thickness, which this article argues is the largest controllable zinc cost, and the by-product percentages get worse while nothing about the by-products has changed.
That objection is correct, and it does not make the percentage wrong. It makes it a different instrument. A percentage of zinc is a composition figure: it answers where the zinc went, and it is the right form for a material balance, a supplier reconciliation, or one line to a finance director. Kilograms per tonne of steel is a control figure: it answers what the process made per tonne of work, against a denominator the process cannot move. "Ash 14%" cannot be posted on a board and checked against today's production. "8.22 kg per tonne" can. Use the percentage against the ceiling. Use kg per tonne to run the plant.
What follows is a case study. One operator, three plants, one year, 30,510 tonnes. It is not a benchmark and it should not be used as one.
| 2025, three plants combined | Total | Per tonne of steel | % of zinc consumed |
|---|---|---|---|
| Steel galvanized, white weight | 30,510.3 t | — | — |
| Zinc added, equal to zinc consumed | 1,801.0 t | 59.0 kg/t | 5.90% |
| Dross removed | 170.9 t | 5.60 kg/t | 9.0% |
| Ash removed | 250.7 t | 8.22 kg/t | 11.8% |
| Zinc on product, by difference | — | — | 79.2% |
How these were built, including what is assumed. Kettle level is held constant, topped up as zinc is consumed, so zinc added equals zinc consumed and no inventory correction applies. The percentages above are contained zinc, converted from weighed stream mass at 95% for dross and 85% for dry ash, the same assays this article uses elsewhere; an 80% ash assay moves the ash share to 11.1% and the product share to 79.9%. The product share is a residual, so it is an identity rather than a measurement — every weighing error in ash or dross lands in it. Plant 3 contributed one commissioning month of 30.7 tonnes and Plant 2 began in August; the figures are mass-weighted and Plant 3 moves every column by less than 0.03. Ash is 250.7 t, not the 268.7 t in my own first cut: that carried 18.1 t of prior-year ash into the 2025 record for traceability, and it is excluded here because it is not 2025 generation.
That 18.1 tonnes is not a bookkeeping curiosity. Crowley's observation, from auditing plants at scale, is that many galvanizers do not count ash and dross until they sell it — at which point the figure is rounded to the nearest container or truckload and lands in whichever year the sale fell. A number built that way measures disposal, not generation, and the two only agree over a long enough run.
His remedy where the records are like that is to average over a long period, up to five years, so the timing error between what a plant produced and what it sold becomes small against the total. It is a blunt instrument, and it is better than comparing two plants on one year of sales data and concluding something about their process.
The underlying point is about who owns the measurement. Ash and dross tonnages are production statistics, and they answer a production question — is the flux right, is the work dry, is the bath under control. Let the accountants measure money, and let production people measure production, so that the number reaches the person who can act on it while it still means something.
| Stream | Case study | Against the ceiling | Against the aim point |
|---|---|---|---|
| Dross | 9.0% · 5.60 kg/t | Well under ~15% | Inside 3–9 kg/t |
| Ash | 11.8% · 8.22 kg/t | Under ~14% | 1.6× the top of 2–5 kg/t |
Both streams clear the ceiling. Only one is anywhere near the aim point. That gap is the whole argument for keeping both denominators: read as a percentage against the limit, my ash looks fine and I would have stopped looking. Read as 8.22 kilograms against a well-run plant's 2 to 5, it is the largest correctable number in my operation, and the three conditions behind it are set out under KPI 4.
The same edge applied to the rest of my own numbers, because a case study that is hard on one figure and generous with the others is not worth reading. Dross at 5.60 kg/t is inside the field range but mid-range within it, and the lever costing later in this article models a controlled plant at 4 kg/t against 8 for an undisciplined one, so 5.60 is acceptable rather than good. Consumption at 5.90% sits at the top of the normal band in my own table, one tenth of a point below the figure at which this article tells the reader to investigate. It is also imported rather than derived: KPI 1 says do not import a benchmark, calculate one, and I have not yet published a surface-area target for my own work mix. On my own instruction, that number is unfinished.
This is the primary zinc efficiency metric. It captures the cumulative effect of every operational decision made during the period: bath temperature, flux maintenance, immersion time, surface preparation, dross and ash management, and weighing accuracy.
Do not import a benchmark. Calculate one. Zinc on the product is fixed by two things you already know: how much surface area a tonne of your steel carries, and how thick a coating you put on it.
Worked example. A shop whose average measured section is 8 mm, delivering an average coating of 115 µm: S = 255 ÷ 8 = 31.9 m²/t; coating mass = 115 × 7.14 = 821 g/m²; zinc on product = 31.9 × 821 ÷ 10,000 = 2.62%. Add 0.5 to 1.2 points for dross and ash and the target total consumption is 3.1 to 3.8%. If that shop is reporting 6%, the gap is not a mystery to be theorised about. It is level correction, weighing, dross rate, or coating thickness, and the reconciliation will tell you which.
Zinc is deposited per unit of surface area, but consumption is expressed per unit of weight. A tonne of 2 mm material carries roughly 128 m² of surface. A tonne of 20 mm plate carries roughly 13 m². The thin material therefore takes an order of magnitude more zinc per tonne, even though its specified coating is thinner. Coating grade rises with thickness under ASTM A123, but nowhere near fast enough to offset a tenfold change in surface area.
| Work category (average measured section) | Surface area | Zinc on product | Total consumption target |
|---|---|---|---|
| Heavy plate & girders ≥16 mm | 10–16 m²/t | 1.0–1.6% | 1.5–2.8% |
| Structural sections 6–16 mm | 16–42 m²/t | 1.5–3.5% | 2.2–4.5% |
| Medium sections 3–6 mm | 42–85 m²/t | 3.0–5.5% | 3.6–6.5% |
| Light sections & tube 1.5–3 mm | 85–170 m²/t | 5.5–9% | 6.2–10% |
| Centrifuge: fasteners, small parts | Very high, variable | 4–8% | Track by part family, never blended |
Across ordinary job-shop plants running mixed work at 10,000 to 75,000 tonnes per year, the observed field range for total zinc consumption is 4 to 6% on a white-steel basis. Plants outside that range exist, but they are exceptional in one direction or the other rather than typical. Use the table below only for a mixed programme, and only after the level correction is applied.
| Blended consumption % | Status | Reading |
|---|---|---|
| < 3.5% | Verify first | Plausible only for genuinely heavy work. Otherwise check level correction, jig tare and weight basis before celebrating. |
| 4.0–5.0% | Strong | Well-controlled mixed programme with disciplined dross and ash management. |
| 5.0–6.0% | Normal | Typical field range for mixed job-shop work. |
| 6.0–7.0% | Investigate | Normal only if the work mix is genuinely light. Otherwise check dross rate, flux concentration and coating thickness control. |
| > 7.0% | Action required | Systemic: level error, weighing error, over-thick coatings, or a dross and ash problem. |
ASTM A123/A123M-24 was approved in May 2024 and published in July 2024, the first substantive revision since the 2017 edition. Its actual content is: minimum average coating thickness requirements for forgings and castings, a definition of "intended use" as distinct from intended appearance, a new appendix clarifying interpretation of Table 1 and how measured steel thickness is determined, clarification that "one coating grade below" is read from Table 2, and new notes covering masking, re-galvanizing and elevated aesthetic requirements.
The new appendix on material-category assignment has a direct and often overlooked effect on required coating grade, and therefore on zinc consumed per tonne. Category is decided by how the product is made, not by what it looks like.
On 12 mm material at 21 m²/t, 25 µm of coating grade is about 3.8 kg of zinc per tonne, or 0.38 percentage points of consumption. A shop that has been galvanizing plate girders to the structural-shape grade has been giving away zinc it was never contracted to supply.
Spinner plants show wide variation by part geometry, and blended monthly averages are close to meaningless. Track consumption by part family (fasteners against brackets against bar stock) and set a target for each. The centrifuge action itself, together with the very high surface-to-mass ratio of small parts, raises both zinc on product and dross generation relative to batch structural work.
Pickup % is the share of zinc consumed that actually ended up on the product doing its job. It is a reconciliation cross-check rather than a control metric, because you cannot act on it directly. When it moves, one of the other four numbers moved first.
Below 75%, dross and ash generation is too high, or dross and ash weights are being overstated. Above 92%, suspect the arithmetic before the process: unrecorded dross and ash removals, an uncorrected kettle level, jig weight left in the incoming steel, or outgoing product weighed with quench water still on it.
Dross is the Fe-Zn intermetallic that forms as iron dissolves into the bath and settles to the kettle bottom. Zeta phase is close to 94% zinc and 6% iron by stoichiometry; as removed, with entrained free zinc, dross usually assays 95 to 98% zinc. It is unavoidable. How much you make is very largely within your control.
Dross and zinc consumption are not independent of each other. Every kilogram of dross you make raises the numerator and the denominator, because the zinc in that dross was itself consumed. The ratio therefore moves less than the underlying problem, and two plants with genuinely different dross performance can report similar percentages. Measure dross against an independent variable instead: tonnes of steel galvanized. That normalises for plant size without introducing the circularity, and it is the number an operator can act on.
The same argument applies to ash. Both are corrected in this revision.
| Factor | Effect on dross formation | Control action |
|---|---|---|
| Bath temperature 450 °C → 460 °C | Approximately +10%, not a doubling | Hold 445–455 °C. Log hourly, not just at shift start. |
| Bath temperature above ~480 °C | Attack becomes linear, dross and kettle erosion accelerate sharply | Treat any sustained excursion above 470 °C as an incident, not a deviation. |
| Flux iron above ~4 g/L Fe | Iron fed directly into the bath on every dip | Weekly flux iron test. Regenerate or peroxide-treat before the limit. Report as g/L Fe, not FeCl₂. |
| Over-pickling and rough surface | More exposed area and faster iron dissolution on immersion | Pull steel as soon as it is clean. Monitor free acid and iron in the pickle. |
| Sandelin-range steel, Si ≈ 0.06–0.12% | Thick, uncontrolled zeta growth | The dominant zinc cost here is coating thickness on the product, not bath dross. Check the mill certificate, shorten immersion, and price reactive steel separately. |
| Infrequent dross removal | Reduces working bath volume and insulates the kettle bottom | Remove on the throughput rule below. Sound the kettle bottom to know what is actually there. |
The claim is common in galvanizing literature and it is wrong at these temperatures. It matters, because it is the kind of exaggeration that gets a technical article dismissed by people who run baths for a living, and because it distracts from where temperature really bites. The Fe-Zn reaction does accelerate dramatically, but the transition to rapid linear attack sits well above the normal window, from roughly 480 °C upward.
Field and experimental figures for kettle wall erosion give the picture that dross rates alone do not:
Both figures are throughput-dependent, and the spread is wide. Erosion is driven by how much steel passes through the melt, not by the calendar, so a rate quoted without a tonnage is only half a number. Crowley's field range makes the point: a plant running over 500 tonnes a day — he was on site at one that set a record of 560 t/day — gets roughly three years from a kettle, while a plant galvanizing 3,000 tonnes a year wears its wall at about 1 mm per year. The 2 to 3 mm/year above is a mid-range figure, not a universal one. Before you compare your kettle life with anyone else's, normalise for tonnes through the bath.
That is not a KPI question, it is an asset-life question. My own baths run 445–455 °C and never leave that window, which is the point: erosion of this kind is not a consequence of operating practice, it is a consequence of losing control of the burners. A fault that runs a bath away toward 550 °C for a shift is a capital event. Log bath temperature hourly and alarm it, and treat the thermocouple calibration schedule as a maintenance-critical item.
Weekly drossing is unnecessary below roughly 10,000 tonnes per year. A plant at 6,000 tpa makes around 115 tonnes of steel a week and simply has not generated enough dross to justify the disturbance, the labour, or the zinc carried out with a thin, wet dross layer. Tying the interval to throughput scales correctly across plant sizes. The two-week ceiling covers slow periods where dross has been made but volume has not accumulated.
Confirm with a measurement, not a habit. Sound the kettle bottom with a probe before each planned removal and log the depth. Knowing how much dross is actually present tells you whether to pull it now, how much to expect, and whether your formation rate is changing. Plants that dross on a calendar without sounding the kettle routinely remove either too little or too much zinc-rich material.
Some plants run bath iron as high as 0.08%. Iron held in solution at operating temperature is invisible on the kettle bottom and cannot be skimmed out. Drop the temperature before drossing and a portion of that dissolved iron comes out of solution and reports to the bottom, where it can be removed with the rest. Post-drossing bath iron in the region of 0.03% is achievable this way.
This changes what the operation is for. Cooling before drossing is not just a zinc recovery tactic, it is bath chemistry maintenance. A bath carrying high dissolved iron will keep precipitating dross at every temperature dip and will keep contaminating coatings with dross pimples until the iron is taken out.
What accumulated dross does and does not do. It does not consume bath zinc continuously, as Revision 1.0 implied. What it does is reduce the working volume of the kettle, insulate the kettle bottom from the melt so that bottom and lower-wall metal temperatures rise and erosion accelerates, and provide a reservoir of particles that can be stirred into the coating as dross pimples. The zinc loss itself happens at the moment of removal, because dross always leaves with free zinc entrained in it. That is precisely why the removal should be planned, cooled and measured rather than done in a hurry.
Ash forms at the bath surface from oxidation of molten zinc, from flux residue reacting as fluxed work enters the bath, and from floating fines. Zinc content depends entirely on which material you are actually weighing. Dry ash skimmed from a dry-flux kettle is largely entrained metallic zinc droplets with zinc oxide and flux residue, and commonly assays above 80% zinc, with published figures running to 90% and above. Wet or spent top flux from a wet-galvanizing blanket is a different material at roughly 50% zinc. If your log mixes the two under one heading, your ash rate is meaningless and so is your recovery invoice.
This is the most consequential error in Revision 1.0, because a plant that acted on it would have increased ash generation and coating defects at the same time, while believing it was saving zinc. If you diluted your flux on the strength of the earlier version, restore it.
Why weak flux makes ash: a correctly concentrated flux leaves a continuous, dry, coherent film that protects the steel surface from re-oxidation before immersion and reacts cleanly on entry. A dilute film is patchy and dries poorly, so the work carries residual moisture and partly re-oxidised surface into the bath. The result is more surface turbulence, more oxidation of exposed molten zinc, more unreacted flux at the surface, and more ash. The secondary cost is worse: incomplete fluxing gives bare spots and rework, and every re-dip is another full dose of zinc.
| Flux type | Target total salt concentration | Notes |
|---|---|---|
| Double salt, ZnCl₂·2NH₄Cl | 400–450 g/L | Standard for most batch plants with a drying oven. |
| Triple salt, ZnCl₂·3NH₄Cl | ~310 g/L | Dries more readily than double salt. Preferred where there is no dryer. |
| Double salt at 200–300 g/L | Too weak | Consistently associated with high ash. Correct it. |
| Mono salt | Not recommended | Rarely justified for general batch work, and usually found running under-concentrated as well. |
| Flux iron | < 2 g/L Fe target, ~4 g/L Fe action limit | State the species. 4 g/L Fe is about 10 g/L expressed as FeCl₂. |
Flux strength gets reported as a percentage, in g/L, as density, in degrees Baumé, and as specific gravity, and the numbers are not interchangeable. Percentage is ambiguous unless the basis is stated. Density-based measures, Baumé and SG included, are distorted by dissolved iron in the flux, so a bath that is genuinely weakening in salt can read as unchanged while its iron climbs.
Report total salt in g/L and iron in g/L Fe. Two numbers, unambiguous, and both directly actionable. Standardise the titration method and record it on the log sheet.
Skimming discipline. Ash must be skimmed clear before every withdrawal. Ash still floating when work comes up through the surface is dragged into the coating, which is a defect and a zinc loss in the same movement. It takes seconds and it is never the step to skip under production pressure. Note that skimming controls where the ash ends up, not how much of it forms. The volume is controlled at the flux tank.
Concentration is the condition this article has covered so far, and it is the one most often wrong. It is not the only one. My own ash sits at 8.22 kg per tonne against a well-run plant's 2 to 5, and when I went looking for why, three more conditions came up. All four are properties of the flux solution and the work entering the kettle, not of the ash after it forms. Ash is a symptom. Every control for it sits upstream of the zinc.
| Condition of the flux stage | Where it should be | What happens when it is wrong |
|---|---|---|
| Total salt concentration | Double salt 400–450 g/L; triple salt ~310 g/L | A dilute tank leaves a patchy film and can generate as much as double the ash. |
| Flux solution temperature | Heated, around 50 °C | A cold tank drains and dries poorly. Work carries more solution and less heat into the drying stage, so the film is still wet at the kettle. |
| Drying before immersion | Dry to the touch, every load | Wet work flashes its water off at 450 °C. That blows flux and zinc off the surface as ash, and it is a safety exposure at the kettle in its own right. |
| Flux solution pH | About 4 | Outside band the flux stops behaving as specified, and the concentration written on the log sheet stops describing what is on the steel. |
What my own data does and does not show. The three conditions above are my diagnosis of my own ash rate, from operating the plants. They are not a finding from the case-study numbers. My workbook records tonnage, zinc added, dross and ash, and nothing else: it holds no flux temperature, no pH and no drying record, so it cannot evidence any of them. Nor is flux temperature or pH on my flux log today, which is the first thing to fix — a condition that is not measured cannot be shown to be the cause of anything.
Three of those four cost nothing but discipline. On my numbers the prize is the gap between 8.22 kg per tonne and the 2 to 5 a well-run flux tank delivers, which is the largest correctable figure anywhere in this article's by-product accounting.
This is the financial expression of everything above, and the number that feeds your price per tonne.
Throughput 500 t. Zinc consumption 5.0% level-corrected = 25,000 kg. Delivered zinc cost SAR 12.00/kg, which at a 15% ingot premium implies an LME reference of SAR 10.43/kg. Dross 5.5 kg/t = 2,750 kg sold at 80% of LME. Dry ash 3.5 kg/t = 1,750 kg sold at 60% of LME.
Gross zinc cost: 25,000 × 12.00 = SAR 300,000.
Dross recovery: 2,750 × 10.43 × 0.80 = SAR 22,946.
Ash recovery: 1,750 × 10.43 × 0.60 = SAR 10,952.
Net zinc cost: SAR 266,102 ÷ 500 t = SAR 532 per tonne of steel.
Dross and ash are quoted as a percentage of the LME zinc price. Your ingot is bought at LME plus a delivery and conversion premium, commonly 10 to 20%. Compare a sale quoted off LME against a purchase cost that includes the premium and you will overstate your recovery income by the whole premium, every month, silently.
Rule: convert both sides to the same reference before you net them. State the LME reference, the premium, and the recovery percentage separately on the report so anyone can re-derive the number.
Zinc saving claims are routinely overstated because they value avoided consumption at full purchase price. The real cost of generating a kilogram of dross or ash is the zinc it carries out, less the price you get back for it. At good recovery prices that difference is far smaller than the usual narrative implies.
| By-product | Zinc content | Sale price | Zinc carried out | Recovered | Net cost per kg |
|---|---|---|---|---|---|
| Bottom dross | ~95% | 80% of LME | SAR 11.40 | SAR 8.35 | SAR 3.05 |
| Dry ash | ~85% | 60% of LME | SAR 10.20 | SAR 6.26 | SAR 3.94 |
| Zinc on product | 100% | Not recoverable | SAR 12.00 | SAR 0.00 | SAR 12.00 |
Three things fall out of that table, and none of them is what the standard zinc-saving article tells you.
Almost every article written about zinc cost in this industry, including Revision 1.0 of this one, spends its length on dross and ash. Work through the arithmetic above and that emphasis is difficult to justify.
| Zinc destination | Share of zinc consumed | Net cost, 500 t/month plant | Recovery offset |
|---|---|---|---|
| On the product | 84% | SAR 252,000/month | None |
| Bottom dross | 10% | SAR 8,400/month | 80% of LME |
| Dry ash | 6% | SAR 6,900/month | 60% of LME |
Dross and ash together carry a net cost of about SAR 15,300 a month on this plant. Zinc on the product carries roughly sixteen times that, and none of it comes back. If you eliminated dross and ash entirely, which is physically impossible, you would move the zinc bill by under 6%.
A plant delivering an average of 140 µm where Grade 100 is specified is fully compliant and is giving away 40% of its coating zinc. Bringing the average down to 115 µm, still comfortably above the minimum with margin for measurement scatter, cuts product zinc by about 18%.
On the 500 t/month plant: 21,000 kg of zinc on product, reduced by 18%, is 3,750 kg per month. At SAR 12.00/kg delivered, with no recovery offset because that zinc was never going to be sold, the saving is SAR 45,000 per month, or about SAR 540,000 per year.
That is roughly four and a half times the entire prize available from halving dross and ash together. It is also, unlike a dross reduction programme, immediately visible in the thickness log you are already keeping.
The boundary condition, stated plainly. None of this means shaving coatings toward the limit. Thickness compliance stays at 100%, and the target is to remove the unintended excess, the microns nobody specified, nobody is paid for, and nobody is tracking. That is a control problem, not a specification problem.
Every rework batch is a second full dose of zinc onto steel the customer paid for once, plus a second kettle cycle and a delay to everything behind it in the queue. On a plant running 80 batches a month, 8 rework batches consume the better part of an extra shift. First-pass rate is a zinc KPI as much as a quality one.
Not 97%. Not 99%. If batches leave with readings below the specified minimum, whether the governing document is ASTM A123/A123M-24 or ISO 1461:2022 (fourth edition, August 2022, superseding the 2009 edition), the plant is shipping non-conforming product, and the commercial exposure is not proportional to the production pressure that caused it. Confirm which edition your contracts cite, because superseded editions still circulate in purchase orders.
The other side of the same metric. Compliance measures the floor. Zinc cost lives in the ceiling. A plant averaging 140 µm where 100 µm is specified is 100% compliant and is giving away roughly 40% of its coating zinc. Track average coating thickness against specification minimum as a companion figure, and manage the gap deliberately: reduced immersion and withdrawal time, better drainage, and steel selection where you have any say in it.
Every zinc cost reduction available without capital spend reduces to one of five levers. All five require process discipline applied every shift, every week.
Take a 500 t/month plant moving from undisciplined to controlled: dross from 8 kg/t to 4 kg/t, ash from 5 kg/t to 3 kg/t, using the net by-product costs of SAR 3.05 and SAR 3.94 per kg derived earlier.
Dross: 4 kg/t × 500 t = 2,000 kg not made × SAR 3.05 = SAR 6,100/month.
Ash: 2 kg/t × 500 t = 1,000 kg not made × SAR 3.94 = SAR 3,940/month.
Combined ≈ SAR 10,000 per month, or SAR 120,000 per year.
Worth having, and worth the discipline. But keep it in proportion: the same plant can recover four to five times that figure from average coating thickness alone, and that zinc carries no recovery offset at all. Fix the by-products because they signal bath and flux health, not because they are where the money is.
Revision 1.0 quoted a figure roughly ten times larger for the by-product case. That was a units error, a monthly saving stated as annual, compounded by valuing avoided zinc at full purchase price with no recovery offset. The numbers above will match your actuals.
Alloy additions are effective but they are a consumable cost, so they belong in a different conversation from the five levers above. Aluminium at trace level suppresses surface oxidation and improves drainage. Nickel at around 0.05% suppresses Sandelin reactivity and is the standard answer where reactive steel cannot be avoided. Bismuth or tin improve drainage and reduce zinc pickup on the product. Each one needs a business case against your actual steel mix, and each one changes your dross chemistry, so brief your dross buyer before you start.
The formulas are straightforward. The inputs are where plants fail.
The monthly report is written for management. The plant floor needs something else: a daily display showing where the plant stands against target right now. The version that works is a physical whiteboard, updated every shift by the supervisor, carrying five numbers.
Five numbers, visible to every operator walking in. The one that is off target tells the shift exactly what to look at, and it names the check: flux concentration in g/L. That is the difference between a display and a control. A board full of green numbers teaches nothing.
The mechanism is simple. When the operator knows his shift's numbers go on that board, he pays attention to bath temperature, to skimming before withdrawal, and to whether the 14:00 thermocouple reading matches the 10:00 reading. The KPI becomes personal rather than abstract.
When I first structured the full KPI framework across my three plants, the most revealing early result was not about zinc consumption or throughput. It was about measurement consistency.
In the first month, Plant 2 came out at 6.1%, close to what I expected. Plant 1 came out at 9.3%, which I did not expect at all. The plant had been running profitably for years and nobody had flagged a problem.
The investigation took three weeks. The cause was not a process failure. Jig weight had been included in the incoming steel weight for years, systematically understating the steel processed and inflating the consumption percentage. Corrected, the real figure was 6.4%.
Two things I would add now that I did not know then. First, 6.4% was still high for that plant's work mix once I calculated a surface-area target rather than importing a benchmark. Second, we had no kettle level correction at that point, so part of what we were calling consumption was bath inventory. Both were found later, and both were larger than anything we changed in the process.
The lesson holds and gets stronger: a high zinc consumption number is the start of an investigation, not a verdict. Verify the measurement before you touch the process. Most apparent efficiency problems are measurement problems wearing a costume.
Revision 2.0 exists because a reader took the time to disagree in detail. Most of the corrections below came with useful input sourced from a wide range of galvanizing plants by Geoff Crowley — more than 30 years of direct plant management and audits of roughly 175 plants over the last four years, a body of field data on dross, ash, flux and kettle erosion broader than any single operator can assemble. He has seen the corrections but not this revision, and nothing here should be read as his endorsement; the article and any error left in it are mine. Technical writing is only as good as the people willing to correct it, and I would rather publish a corrected article than a comfortable one.
For the avoidance of doubt, the reviewer and the consultant whose field figures are cited in the benchmark table are the same person. The kg per tonne figures in this article are his, and I have not dressed one source up as two.
The reading of the published percentage benchmarks as ceilings rather than targets is my own position, argued from our realised by-product prices and from our own plant data. It is a disagreement with widely used industry figures, not with the organisations that publish them, and it is offered for exactly the same treatment I received: check the arithmetic and tell me where it fails.
