Plant notes · Bao'an
What actually failed on the line, then what we locked
Named logos stay off this page unless the buyer signs. These are anonymized jobs from the WYD floor (since 2006): soft trays often from MOQ 20, prototype usually 3–7 days after CAD lock, reply typically within 12 hours; complete files help us identify the next steps.
What each case note shows
These notes are written from production records, drawings, photos or inspection information that can be shared without exposing customer identity or protected geometry. Each case should state what the project started with, what had to be cast or manufactured, the manufacturing route, the main risk or observed failure, the design or process change, how the prototype or production result was checked, and which facts are measured versus still confidential.
An anonymized case is not a fictional case. If a result cannot be supported by a project record, it is an engineering note rather than completed customer work. Hub cards below are grouped by job, not by a filter UI.
Buyer note
“I did. I like some of them a lot. The really small ones and the cubes I like. The hearts are too shallow, but that's my fault. I didn't look at the specifications. I'll probably be ordering some more. Thank you.”
Ari · LinkedIn message to Kyler Yang · gummy silicone molds
Cavity depth is a written spec. If a heart, cube or mini piece must hit a fill height, lock that millimetre on the RFQ before the next tray - do not guess from a photo of an empty cavity.
ILY hand: confirm the 3D, then three sheets
Fail: A concept sent straight to silicone comes back with the wrong finger thickness.
Lock: Dimensioned 3D first (A 6.93 / B 10.07 / C 13.05). Then three sheets: 205 × 185, 176 × 110, 230 × 145 mm.
Handshape wax melt →Dragon, tiger, phoenix relief wells — plaster and ointment crafts
Fail: A named three-shape set still fails if one cavity depth is copied onto hair, scale, and tail.
Lock: Stills of wells and plaster-style pulls. Same class can pour large chocolate or freeze large ice if the RFQ names that pour.
Relief set clips →Cacao-fruit chocolate tray, 90 x 35 x 20 mm, Shore A 10
Fail: Organic STL still had mesh bumps. Height and softness were still arguments, not a locked envelope.
Lock: Cleaned STL, locked 20 mm height and 90 mm length, gypsum then silicone sample, 12-piece paid run. Cavity count unpublished — not locked.
Chocolate cacao tray →Four gummy designs: unused July quote, paid August mix
Fail: First assortment was technically possible and still too wide to purchase.
Lock: Revised four-code mix 15 Aug 2026, paid and shipped ~5 kg parcel. Grams/Shore not reverse-engineered from photos.
Gummy requote →Gamepad dome Rev 4, 0.05 mm hole stack, Shore 30/40/50 lots
Fail: Older STEP could have been tooled while the PDF said Rev 4. Production Shore was still a guess.
Lock: Matched Rev 4 PDF+STEP, 10 pcs each at 30/40/50. MOQ 5,000 is this program only. Production Shore still open.
Gamepad dome →Silicone cup: internal lip ridge before silk-screen
Fail: Print on a cup whose internal edge still caught the lip.
Lock: 1-cavity steel sample, Shore 50/60 then 70 feel lots, print after the ridge was gone. Photos held (NDA).
Coffee cup OEM →Ear-cap: first sample fail, second set fit
Fail: First molded set missed the mating product. STL was not enough for a 1 mm lip.
Lock: STEP as tool reference, black Shore A 30, second sample cleared fit. Photos held (NDA).
Ear-cap accessory →Epoxy RTV: inner/outer STEP families, quotation stage
Fail: Pricing 'a resin mold' while wall and height were still missing.
Lock: Paired inner/outer STEP, size matrix, matte vs gloss on the quote. Not a completed production claim.
Epoxy RTV families →Concrete spacer RTV, Shore A 50, plaster before concrete
Fail: Cavity count treated as free output on a heavy sheet.
Lock: Size families 20/25/40/50, example maps 72/30/20 and 150/130/70, plaster demold first.
Spacer RTV →Architectural pads 75 / 50 / 35 x 40, CAD review only
Fail: Three sizes quoted with drifting 2 vs 2.5 mm walls.
Lock: PDF to shop STP, one family rule. CAD evidence — not a shipped production claim.
Pad RTV family →9.2 g / 6.2 g / 8.8 g trays, 70–100 cavities
Fail: New SKUs kept changing cavity count before the single-well grams were frozen. Outer rows and a packed 100-up map would have multiplied the miss.
Lock: Locked each weight family first, then 70 / 80 / 100 maps, Shore A 30 on approved notes, 2023–2026 revision trail — anonymized U.S. line.
Gummy tray case →30-up mantis shrimp tray, 450 × 250 mm
Fail: Organic 皮皮虾 ridges and tail fan die if you nest from a phone photo. No written bbox, no 6 × 5 layout.
Lock: Scan bbox locked at 31.493 × 85.087 × 12.800 mm, then 30 cavities on a 450.00 × 250.00 mm sheet — translucent tray plus process clip.
Scan to mold →Pectin line, outer cavities ran light
Fail: Center pieces hit label grams. Outer row sat 5–8% light once the tray flexed on the depositor.
Lock: We locked mean cavity ml first, then thickened the frame and re-weighed outer vs center before anyone talked steel.
Gummy trays →Photo-dense tray vs nozzle pitch
Fail: Buyer sent a packed cavity photo. Nozzle map was already frozen. Nesting from the photo would have missed the teach points.
Lock: Pitch and nest went on the RFQ before cavity count. Soft sample had to prove fill on the real depositor coordinates.
Depositor trays →Soft sample signed, steel PO same archive
Fail: SKU was still moving weekly, but the line already wanted metal because trays were wearing mid-shift.
Lock: We kept the same CAD pack: signed soft weight and logo depth, then steel on this floor — no second factory redraw.
Soft to production →Kitchen handle overmold, insert walk
Fail: First EVT grips looked fine by hand. Inserts walked once the press cycle ran hot.
Lock: Bond face and insert location were measured against the mating part, not a mood board Shore A.
Overmolding →Cavity ml → grams: cavity calculator. Quote fields: RFQ builder.
If your line already has a fail, send that — not a pretty tray photo
Outer weight, pitch, insert walk, or a sample that does not match production Shore. We quote the path from Bao'an.