Warehouse and WIP control: parts between operations

A pressed bracket that waits nine days between the press and the paint booth arrives at pretreatment with rust on every cut edge. Nothing was done to it. It sat in a stillage, in a monsoon, and the clock did the damage. The same bracket moved in three days coats without a derust stage, and the difference is WIP age, not process capability.

WIP control is usually managed as a scheduling question — how much buffer absorbs a breakdown, how much stock protects a customer release. It is also a quality variable, and a more measurable one than most. Every day a batch spends on the floor is another day of handling, another day of humidity, and another chance for a superseded revision to reach the paint line.

The arithmetic is simple enough to run in a meeting. WIP equals throughput multiplied by the time a part spends in the route. Everything else on this page follows from shortening that time and protecting the part while it waits.

How much WIP should WIP control hold?

Little's Law applies directly: average WIP equals average throughput multiplied by average cycle time. A plant processing 200 panels a day with a six-day route from laser to dispatch is holding 1,200 panels on the floor at any moment.

That number is not abstract. At 0.6 m² per panel it is 720 m² of painted or bare steel, stacked 50 high, which is at least 24 stacks and roughly 18 m² of floor plus the aisles to reach them. It is also 1,200 panels that will each be handled four to six times before dispatch — around 6,000 touches a day, each one a chance to dent a corner.

Cut the route to three days and the WIP halves to 600 panels. Nothing about the machines changes. The handling exposure halves, the rust window stops being marginal, and the floor space released is enough for another cell.

The four mechanisms that damage parts between operations

Compression. Panels stacked flat under their own weight take a permanent set across an unsupported span, and the bottom panel takes the most. A 0.8 mm panel spanning 500 mm with 100 panels above it deflects visibly, and it does not spring back flat. The fix is to stack on edge, limit the stack height, or support the panel at a stiffening feature.

Abrasion. Two parts sliding against each other remove coating and scratch bare steel. It happens when a stillage is dragged, when a part is thrown onto a pile, and when a stack is unstrapped and the parts shift. Interleaving and part-specific dunnage stop it; better housekeeping does not.

Point loading. A bent panel hung on a single hook deforms at the contact point, and the deformation is exactly where the customer looks. Wide hooks, slings or a rack with two supports spread the load.

Contamination. Dust, grinding sparks, oil from a forklift mast and weld spatter from the next bay all land on parts waiting for paint. A covered rack is not bureaucracy; it is the difference between a part that passes the water-break test and one that does not.

Rust sits across all four. It is a chemical consequence of time and humidity, and it makes every other damage mode worse because a rusty surface has to be blasted or re-cleaned, which adds handling.

Racking, dunnage and the cost of a bad stillage

The default storage in most sheet metal plants is a stillage with a plywood base and four posts. It is cheap and it is wrong for anything painted, because it stacks parts flat and lets them rub.

Three storage types cover most of a plant. Cantilever racks hold long sections and bent panels vertically, which is the orientation that drains, dries and protects a formed part. Stillages with dividers hold flat blanks and pressed parts in counted stacks. Bins hold small brackets and hardware, and they are the worst option for anything with a cosmetic face because parts tumble against each other.

Dunnage earns its place on three categories: large panels, painted parts and parts with a critical flatness callout. For everything else, a standard stillage with a divider set and a stated stack height is enough, and the discipline of writing the stack height on the stillage matters more than the stillage itself.

The test of a storage method is what happens when someone is in a hurry. If the fast way to move a batch damages it, the batch will be damaged on the shift when the truck is waiting.

The rust window and humidity control

CRCA arrives oiled for a reason, and every operation that removes oil shortens the time the part can wait. Cutting removes the oil at the edge. Degreasing removes it everywhere. A part that has been through pretreatment and not yet coated is the most vulnerable item in the plant, which is why the 24-hour limit before the booth is the tightest number in the table.

Humidity is the variable to watch. Under 60% relative humidity, a clean CRCA part is stable for days. Between 60% and 75% the window shrinks to about 48 hours. Above 85%, a laser-cut edge rusts overnight, and a monsoon week makes the whole plant a corrosion chamber.

Three controls work. Keep the WIP store enclosed and ventilated rather than open to the yard. Keep bare parts covered, with the cover off the floor so air moves. And schedule the oldest batch into pretreatment first, every shift, so nothing sits long enough to reach the limit.

Where a part must wait longer, the options are oiling, VCI paper interleaved in the stack, or a desiccant in a sealed pack. Each adds a handling step; shortening the wait usually costs less.

Tagging, FIFO and revision control

A batch without a tag is not a batch, it is a pile. The tag on a stillage has to carry five fields: part number, drawing revision, quantity, the last operation completed, and the date that operation finished. Everything else — customer, order number, material heat — can be on the route card that travels with it.

The date is the field that makes FIFO possible. Without it, the oldest batch on the floor is identified by memory, and memory always picks the nearest stillage. With it, the store can walk the row and pull the oldest date into the next operation, which is what keeps the rust window and the revision control working at the same time.

Revision control fails in a specific way: a change is issued, the new drawing reaches the laser, and a stillage of the previous revision keeps moving down the line because nothing on the tag says which revision it is. It arrives at paint, gets coated, and the error is found at assembly. A revision field on the tag, checked at every operation, closes that gap for the cost of a printed label.

Two more practices make the system visible. Colour-code the tag by week of entry, so a stillage that is three colours old stands out from across the aisle. And use a red tag for anything on hold, with the reason written on it — a red tag without a reason is a stillage that will be moved by the next person who needs the space.

Kanban and the supermarket between operations

A supermarket is a controlled buffer of standard parts, restocked by the downstream operation pulling what it needs. It works between laser and bend, between bend and weld, and between weld and paint, because those are the points where a breakdown upstream would otherwise stop everything downstream.

Size the buffer from demand during replenishment, not from floor space. If the bend cell consumes 200 panels a day, the replenishment loop takes one day and the plant wants half a day of protection, the supermarket holds 300 panels. That is the whole calculation, and it is the number that should be written on the rack.

The two-bin method is the simplest version and the one that survives a busy shift. Two containers of the same part sit at the cell; the operator works from one, and when it empties, the empty bin goes back as the replenishment signal while the second bin is consumed. There is no card to lose and no software to update, and the buffer size is set by the bin size.

A supermarket is not free floor space. It is WIP with a purpose, and every panel in it is exposed to the same handling and humidity as any other. The rule is that the supermarket has a maximum, and a full supermarket means the upstream cell stops for a while rather than producing more.

Counting: the number nobody trusts

Most plants cannot state their WIP count to within 10%, and the error is invisible until a release is short. The cycle count is the control: count a defined set of stillages every week, compare with the recorded quantity, and track accuracy as a percentage.

Accuracy is counted quantity divided by recorded quantity, averaged over the sample. A plant at 85% is making decisions on numbers that are wrong one time in seven, which produces both shortages and excess. A plant at 98% can run a supermarket without a safety stock on top of it.

The causes of error are predictable: parts removed for a sample and not recorded, two stillages merged into one, a partial batch split at a machine, and rejects returned to the pile without a quantity. Each is a one-line fix at the point where it happens, and none is fixed by a stock system alone.

Count accuracy also feeds the damage question. A stillage that has been opened, partly consumed and left uncovered is both miscounted and exposed. The count and the condition check belong in the same walk.

What Ganesh runs

Ganesh has manufactured sheet metal components since 1992 across five units in Mumbai (Vasai) and Nagpur, with mechanical presses from 20 T to 300 T, a 12 kW fibre laser with a 40 ft × 16 ft bed, a 200 T 5-axis brake with a 4,450 mm bed, three CLOOS robotic MIG cells and more than 30 manual MIG, spot and SPM stations. Parts move between those operations on stillages, racks and bins, and the WIP rules are written into the route card rather than left to the shift.

The plants hold ISO 9001:2015 and ISO 14001:2015. Under ISO 9001 the identification and traceability clause is what makes the tag fields mandatory: a batch has to be identifiable at every stage, and a nonconforming batch has to be segregated and marked. The red tag and the revision field are the visible part of that requirement.

Long parts present the hardest storage problem, because a 4,450 mm bent section cannot be stacked and cannot be hung from one point. The long-bed bending guide covers the forming side, the packing and dispatch route covers what happens once the batch is ready to leave, and the plant and quality hub collects the traceability and storage records behind both.

What to put in the supply agreement

WIP rules are usually unstated, which is why they break under pressure. Put these in the agreement and the auditor has something to check, and tie them to the release pattern in recurring orders and capacity planning.

• Maximum WIP age per operation, and the derust or re-clean route if the limit is exceeded.
• Storage method for each part family: stillage, vertical rack or bin, with the stack height.
• Interleaving and dunnage requirements for cosmetic faces.
• Humidity limit for the WIP store, where the part is bare steel.
• Tag fields and the revision check at each operation.
• Buffer sizes at the named supermarkets, and who owns the stock in them.
• Cycle count frequency and the accuracy target.
• The escalation route when a batch is red-tagged.

Those eight lines do more for incoming quality than most of the inspection clauses in a typical purchase order.

When a supermarket is the wrong answer

A supermarket adds handling. For a part that is made and consumed in the same shift, the buffer is a stillage that gets in the way, and single-piece flow from laser to bend to weld is faster and gentler.

The same applies below about 20 active part numbers on a stable schedule. With few parts and predictable demand, the route can be sequenced directly, and the buffer only hides the fact that the schedule is not being followed.

Large parts break the model altogether. A 4,450 mm section needs a rack, a wide aisle and two people to move, and a supermarket of ten of them is a warehouse, not a buffer. Those parts are scheduled to order, stored vertically, and moved once.

Where the plant runs a genuinely mixed model with short cycles, the supermarket is the right tool and the buffer sizes above are the ones to use. Where it does not, the shortest route with the fewest touches is the quality control.

Shorten the route, tag the batch with a date, and store the part in the orientation it was designed to keep. WIP then stops being the place where good parts go bad. Send the drawing, volume and release pattern through the RFQ and the quotation states the buffer and pack.