How to Calculate a Stainless Fabrication Delivery Schedule You Won't Miss: Critical Path, Welding Hours and Truck-Load Math

When custom stainless steelwork — balustrades, grilles, manhole covers, curtain-wall trims, decorative profiles — arrives late, the cause is rarely that the factory lacked capacity. It is that the schedule was mis-estimated from day one. Three errors account for most of it: quoting the fastest operation as if it were the whole job, forgetting that drawing approval consumes calendar days, and treating transport as a one-day formality.

This article sets out a chain you can copy: critical-path nodes → welding hours back-calculated from seam length → queue time from Little's Law → truck count from a two-gate formula. A 500-unit stainless balustrade order is worked end to end, and every number can be re-run with your own parameters.

Step 1: Eight nodes, and which ones can run in parallel

Delivery is a network, not a line. Anything that can overlap must overlap; only the serial portion actually consumes the calendar.

Theoretical critical path = 2 + 7.5 + 1.5 + 8.3 + 1.5 + 2 + 2 = 24.8 days; with 10% buffer, 27.5 days. If material has to be mill-ordered (non-standard sizes or special grades), node 3 grows from 2 to 20 days and the total jumps to 45.5 days.

Both big variables sit outside the workshop. Drawing approval and material arrival together account for 38% of the standard-order cycle and 60% of a mill-order cycle — and they are precisely the two steps the fabricator controls least and the client controls most. One day saved in signing off drawings is worth more than two extra shifts in the shop. That is why we list both as separate client-action milestones in the contract appendix.

Three overlaps that actually save days

• 2 with 3 — release stock sizes against estimated quantities, then top up specials after approval. Saves 5–7 days
• 5 with 6 — start the finishing line once 60% of welding is done instead of waiting for the last seam. Saves 1.5–2 days
• 7 in batches — pack as batches complete rather than waiting for the full order. Saves 1–1.5 days

All three together bring the standard order to roughly 20 days. The cost is extra floor space and more work in progress, so it is worth doing only where the delay penalty justifies it.

Step 2: Derive capacity from welding hours, not from "units per day"

Estimating output as "how many sets we build in a day" is wrong because seam length varies several-fold between projects. Work backwards from metres of weld:

• v_weld — travel speed. Manual TIG on 3 mm-thin 304 fillet welds: 100 mm/min (thin plate 80–150, thick plate down to 40–60)
• eta_arc — arc-on ratio, the share of shift time the arc is actually burning. On manual fabrication floors this measures 0.35–0.45; the rest is fit-up, clamping, repositioning, grinding and waiting for the crane
• Effective hours — 8 h shift minus start-up and clean-down: 7.5 h

Worked example: 500 units, 6 m of weld each

The arc-on ratio is the biggest lever in this table. Moving eta from 0.40 to 0.50 drops the duration from 8.3 days to 6.7 — equivalent to adding 3.3 welders for free. And you raise eta not by asking people to weld faster but by giving them jigs, material racks and crane scheduling so they do not have to stand up, which is far cheaper than hiring. Most fabricators who cannot quote a reliable date have simply never measured their own eta.

Do not forget finishing, though it is rarely the bottleneck

A flat brushing line feeding at 3 m/min over a 1.2 m effective width at 65% utilisation:

500 units need about 1800 m² treated, i.e. 1.6 shifts — not a bottleneck against 8.3 days of welding. The real queue risk is a paint curing oven: at 30 min per load and 40 posts per load, a shift yields 640 posts; 500 units × 4 posts = 2000 posts needs 3.1 shifts. Once fluorocarbon coating enters the schedule, count oven loads, not square metres.

Step 3: Little's Law for the queue you forgot

Everything above is the theoretical duration of one isolated order. In reality a dozen orders share the floor and yours waits in front of each operation. This is the missing piece behind "you said 25 days, it took 40".

At 800 units per month over 22 working days, R = 800 / 22 = 36.4 units/day:

Use it backwards: to land under 30 days, queue time must stay below 5.2 days, which means WIP under 190 units. Lead time is a simultaneous equation in capacity and work in progress — chasing the order does not solve it. The fuller the floor, the slower every single order moves.

A practical check for buyers: when you visit to review progress, do not only ask "how far along is my order" — ask "how much WIP is on the floor right now". Above roughly 70% of monthly capacity (560 units in this example) the schedule has slipped out of control and any date quoted should be discounted.

Step 4: Truck count — weight and volume are two separate gates

The classic shipping error is counting tonnes only. Long steel members are volumetric cargo; volume tops out long before weight does.

• G_limit — 31 t for a 13 m semi-trailer; 26.5 t of cargo for a 40HQ container
• V_truck — usable stack volume on a 13 m semi: 13.0 × 2.35 × 2.40 ≈ 73.3 m³ (4 m overall height limit, deck included); a 40HQ gives about 68 m³ usable
• eta_load — loading factor: 0.85 for regular cartons, 0.70–0.80 for long irregular members that cannot be stacked and need strapping access

Worked example: 500 units at 60 kg and 0.136 m³ packed

Only 5.6% of what you ship is steel; the rest is air. One unit is 60 kg ÷ 7930 kg/m³ = 0.0076 m³ of metal, yet occupies 0.136 m³ packed — 94% of the freight bill is void space. So the cheapest way to cut transport cost is rarely negotiating the rate, it is repacking: stand posts upright in a demountable frame and fill the voids between them with rail bundles and fitting cartons, lifting eta from 0.75 to 0.82, which turns 2 trucks into 1.5. On large orders that beats any discount.

Step 5: Split shipments by the "smallest installable unit"

Delivering everything at once is often worse than useless on site: no storage, damage from following trades, lost fixings. Batch against the installer's rhythm instead:

• First batch = smallest installable unit — everything needed for one typical floor, one fire compartment or one stair flight, including fittings and fasteners, so a crew finishes a complete section instead of leaving half-work
• Batch interval = the crew's per-section duration, typically 3–5 days; do not ship faster than the site can absorb
• Pack vulnerable items separately with 3% spare — glass channels, decorative covers, fasteners. Losing one stops the crew and costs far more than the spare stock
• Hold 5% of the last batch as a spares pool for site cutting errors and transit damage. Waiting until the end and re-ordering from the factory is a 15-day minimum round trip

Five mistakes that keep repeating

1. Quoting the fastest operation — using 1.5 days of cutting to represent a process actually governed by 8.3 days of welding plus queue
2. Excluding drawing approval — three rounds at 2.5 days is 30% of a standard order. Put a client response time in the contract
3. Assuming stock material — standard 304/316 in standard gauges is stock; non-standard sizes, special grades or large tonnages are mill orders at 20 days, which doubles the schedule
4. Planning manual welding at theoretical travel speed — forgetting the 0.40 arc-on ratio means real hours are 2.5× the estimate; this is the number one reason shops cannot quote a date
5. Counting tonnes but not volume — long members are volumetric, eta is only 0.75, and a weight-based truck count will always come up short

In one line: delivery time = critical-path node days + duration back-calculated from capacity-limited operations + queue time from Little's Law. Compute them separately, buffer each, and you will beat any round "about a month". To compress it, work on drawing approval, early material release, arc-on ratio and loading factor — all four beat overtime, and all four are cheaper.