Blanket Orders and Kan-Ban Scheduling for Custom Silicone Component Supply Chains

A blanket purchase order is an envelope: SKU, price structure, quantity cap, term. Kanban is a pull signal: a card, bin, or EDI message that authorizes the next lot. Neither one is a forecast, and neither one makes silicone behave like a bin of bolts.

Custom LSR, HCR, and extrusion still have mix or mill time, heat cure, post-cure, lot identity, and (for some compounds) a use-by date on the gum. If you staple a two-bin card to a 4-cavity LSR tool with a six-week overseas transit, the card will look busy while the line starves.

Write the envelope and the signal as a system. Then put the chemistry and the tool in the system, not the other way around.

Blanket PO: what it must freeze, and what it must not

Freeze identity: part number, revision, compound (LSR kit or HCR batch spec), color, post-cure, flash spec, packaging, quality agreement. A blanket that says “silicone gasket, 50A” is a license to substitute.

Freeze commercial structure: unit price *by release size* or a stated price break table, tooling already paid and owned, validity dates, currency, Incoterms, tariff/HS handling. Price that assumes 10,000-piece drops and then gets used for 400-piece kanban pulls is not a price. It is a fight.

Cap quantity and time: maximum pieces or maximum value, expiry. A blanket with no cap is an open liability for the shop and a surprise for your finance team.

Do not freeze the shop’s mix of presses as if it were inventory. Capacity is a forecast conversation. The blanket should say how you will share a rolling forecast (even a bad one) and how the shop may buy gum against it.

Index or reopen raw material. Siloxane and platinum LSR kits move. A 12-month frozen unit price with no material trigger is how shops pad the first quote or how they disappear in month seven. A trigger tied to a named index or to documented mill invoices is a structure. A handshake “we will be fair” is not.

Kanban that matches a thermoset cell

Classic two-bin kanban assumes the replenishment time is shorter than the consumption of one bin, and that the item is interchangeable. Custom silicone breaks both assumptions unless you design for them.

Loop time, honestly. Queue at the press + cure-dominated cycle × quantity / cavities + post-cure + inspect/pack + inland + ocean or air + your receiving. For overseas LSR, loop time is often weeks, not the afternoon a Toyota card expects. Bin quantity must cover loop time × consumption, plus yield loss and a quality-hold buffer. If that bin is larger than the tool’s honest batch, you are not running kanban. You are running a min-max warehouse with extra cardboard.

Lot identity travels with the card. One kanban card should not authorize mixing two LSR lots in one carton if your device file forbids it. Print lot, tool number, and cure date on the traveler. FIFO in your cage, not “open the pretty bag first.”

Shelf life is a kanban constraint. Some platinum LSR kits and mixed HCR batches have manufacturer use-by dates. Finished parts can take a compression-set or surface age you care about even when the gum is stable. A card that pulls “whatever is oldest at the factory” can deliver aged stock you never ordered. State maximum age at ship and at receipt.

Setup and color. LSR color changes and HCR compound changes are not free. Kanban that chatters between colors will pay in purge and start-up scrap. Group cards, or pay for a dedicated cavity set.

MOQ, tool occupancy, and the lie of “pull anything”

The press does not care about your card. A 4-cavity Class 103 tool that needs a two-hour heat-up and a 20-shot start-up does not want a 50-piece pull. Either:

• set the kanban quantity at or above a stated runnable lot (start-up scrap included in the economics), or
• accept that the shop will batch multiple cards, which adds delay you must put in loop time, or
• pay a setup surcharge for true small pulls, in the blanket, in writing.

RTV vacuum-cast parts can look more kanban-friendly (no steel heat-up) and still fail: jacket availability, technician calendar, and platinum inhibition do not pull like a bin.

Extrusion is a third animal: die in the press, line speed, start-up footage as scrap. Kanban on cut lengths should authorize a run, not a handful of pieces that force a die change twice a week.

Contract terms that make the pull real

Forecast vs firm. Rolling 12-week forecast, last 2-4 weeks firm (numbers depend on loop time). Kanban consumes the firm window. The shop may buy gum against the forecast at its risk unless you agree to a raw-material buy-ahead with a restocking rule.

Gum vs parts. If siloxane lead times dominate, the blanket may authorize the shop to hold *named* LSR kits or HCR lots for your SKU. That is not consignment of finished goods. It is a raw hedge. Write who owns it, who pays if the revision dies, and the use-by date.

Failed cards. If a pull cannot ship (tool down, yield crash, gum short), the signal is not “try harder.” It is a written escalation: substitute lot only with approval, air freight at whose cost, temporary RTV or 1-cavity aluminum only if that process is qualified.

Inventory location. Factory hold, forward warehouse, or your cage. Kanban from a factory hold overseas is a long loop. Kanban from a local hub is a short loop you paid to create. Do not pretend they are the same card.

Revision control. A drawing change kills the bin. The blanket should freeze how WIP and finished cards are reworked or scrapped, and who pays. Kanban without revision rules will ship the old lip because it was in the full bin.

Failure modes to walk in the kickoff

• Forecast optimism filling the factory cage with your unique 30A LSR in a color nobody else will buy.
• Cards sized to weekly use while ocean transit is monthly.
• Piece price from a 10,000-break used on 300-piece pulls.
• Mixing medical lots to “make the bin.”
• No PM window: the tool never comes out because every week has a card, until the shutoff flashes in a rush lot.
• Dual-source cards with no qualified second tool, which is a single source with extra PDF.

Kickoff test: pick one SKU, write loop time from a timed process plus real transit, size the bin, size the runnable lot, and see if they are within a factor of two. If the runnable lot is 2,000 and the bin that matches consumption is 200, you do not have a kanban design. You have a meeting. Fix lot size, hub inventory, or consumption. Do not print smaller cards.

FAQ

Is a blanket order the same as a forecast?

No: A blanket is the commercial envelope (SKU, price structure, cap, term). A forecast is a non-binding (unless you say otherwise) look-ahead the shop uses to buy gum and slot presses. Kanban is the firm pull inside that envelope. If you skip the forecast, the shop will either over-buy unique compound or miss the first card.

Can I kanban custom LSR the way I kanban catalog O-rings?

Only if loop time, runnable lot, and lot-trace rules fit a two-bin story. Catalog rings often sit in a local DC. Custom 4-cavity LSR with overseas transit does not. You may still use cards from a hub you stocked on a blanket. The factory cell itself is usually min-max or MRP with a kanban face.

Who owns leftover compound bought against my forecast?

Whoever the blanket says. Default shop practice is: they own gum they bought, and they will push you to take it if the revision dies. If you want dedicated kits reserved, write a buy-ahead with a use-by date, a restocking or take-or-pay rule, and a ban on substituting that lot onto another customer without consent.

How do I set kanban quantity without fake weekly usage?

Measure consumption, then set bin ≥ loop time × consumption × (1 + planned scrap) plus a quality-hold buffer. Compare that bin to the factory’s runnable lot. If they disagree, change the network (hub, dual bins, larger releases), not the arithmetic.

Does kanban remove the need for a quality agreement?

No: Pull systems ship faster. They do not inspect for you. Keep FAI on revision, incoming AQL or skip-lot as agreed, lot tickets, and change notification. A card is an authorization to make, not a waiver of the flash spec.