Views: 0 Author: Tonney Shao Publish Time: 2026-08-06 Origin: Abery Mold
In one sentence: In 2011 we finished a two-part injection mold project for a UK Tier-1 electronics supplier whose end customer was JCB — a PBT+15% GF front panel overmolded with Molex connectors and a Nylon+33% GF box body — and although our sample parts passed end-customer sign-off with the help of assembly fixtures we had used to compensate for glass-fiber anisotropic warpage in the box, the first 5,000-piece production run — molded without those fixtures at scale — was rejected in the field, and after fifteen years of thinking about what we could have done differently, my honest conclusion is that we should have refused to submit fixtured samples for sign-off in the first place, we should have insisted on a design revision adding internal ribs to the box body before we cut steel, and because we did not do those two things, we absorbed the full cost of the remake — approximately USD 10,000 — even though contractually the signed-off samples released us from that obligation.
Item | Detail |
Direct customer | A UK Tier-1 electronics supplier (identity withheld under NDA) |
End application | ECU housing family destined for JCB construction equipment |
Year | 2011 |
Parts | (a) Front panel — PBT + 15% GF, overmolded with Molex connectors · (b) Box body — Nylon + 33% GF |
Root technical issue | Nylon 33% GF has 3-5× anisotropic shrinkage between flow and transverse directions; the box body geometry had no internal ribs to resist the resulting warp on the opening flanges |
What we did during sampling | Used assembly fixtures ("jigs") to hold the box body dimensionally correct at inspection; samples passed end-customer sign-off in that condition |
What happened in mass production | 5,000 parts molded without the sample fixtures; opening-flange deformation returned; end customer rejected the entire batch |
Our commercial position | Contractually clean — signed-off samples released our obligation |
What we did anyway | Remade 5,000 parts at our own cost, air-freighted by DHL express |
Financial impact | ~USD 10,000 total loss on the project |
What the customer did next | Placed a significantly larger order with us a few months later — but that outcome is not the point of this case study, and I explain why below |
Above: the box body geometry that this case study is about — a Nylon+33% GF ECU housing with two large connector apertures, an auxiliary aperture, and open perimeter flanges. The flat, un-ribbed inside surface visible in the render is what allowed anisotropic shrinkage to translate directly into flange-plane warp.
I have already published four case studies on this site. Three of them describe things we did well. One — the Russian slider mold we remade for free in 2021 — describes an engineering mistake we owned and paid for. This case study is the second one in that category, and it is the older mistake. Fifteen years old. It taught our engineering team a real lesson about how to say no during DFM — and I think a serious buyer evaluating Abery Mold for a program on a certified end-application like a JCB ECU deserves to see how we think about that class of decision, not just the projects where everything went right.
I am also publishing it named, at customer level, because the end-application manufacturer — JCB — is a global brand that anyone in construction equipment recognizes, and I do not want the technical lesson in this article to be blurred by unnecessary anonymization. Our direct customer (the Tier-1 supplier) remains anonymized under our NDA with them.
The project was a two-piece injection mold program for an ECU housing — the plastic enclosure that protects an Engine Control Unit or a similar control module inside a construction-equipment machine. The two parts were:
Front panel — geometry dominated by two large multi-pin Molex connector openings and an auxiliary aperture. Material: PBT + 15% GF, a standard flame-retardant-capable engineering thermoplastic used across the electrical-connector world for exactly this reason. The connectors were overmolded, meaning the Molex housings were placed into the mold as inserts and the PBT was injected around them, creating a single sealed part with the connectors permanently integrated into the front face.
Box body — the main structural enclosure that mates with the front panel to close the ECU. Material: Nylon + 33% GF (typically PA6-GF33 or PA66-GF33), chosen for stiffness, heat resistance, and dimensional strength under vibration loads that are severe in construction-equipment applications.
Both molds were designed and built in our Shenzhen facility. The front panel came out well: PBT+15% GF has moderate, relatively predictable shrinkage, and overmolding onto Molex connector housings is a discipline we had been doing since the year we started. That side of the project was clean.
The box body was where the problem started, and I want to describe it precisely.
Glass-fiber-reinforced nylon does not shrink evenly. This is not an Abery-specific insight — it is written in every polymer supplier's design guide and in every mold-flow textbook — but it is the single most important thing to understand about the failure I am describing.
When you inject a 30–33% glass-filled nylon compound into a mold, the fibers align in the direction of flow as the polymer front advances through the cavity. Once frozen:
Shrinkage in the flow direction is dramatically reduced by the fibers, typically to around 0.2–0.5%
Shrinkage perpendicular to flow is barely affected by the fibers, typically 0.5–1.0%
That is a 3-5× ratio between flow and transverse shrinkage. On a geometry with balanced flow paths and generous internal reinforcement — internal ribs, gussets, thick-to-thin transitions handled correctly — the difference can be managed and the finished part comes out flat. On a geometry that is a shallow open box with un-reinforced flanges around the perimeter, the differential shrinkage translates directly into flange-plane deformation. The perimeter of the opening tries to shrink harder in one direction than in the other, there is nothing structurally holding it flat, and the flange bows.
That is exactly what our box body did. And the reason it did so is not that we had picked the wrong material — Nylon+33% GF was correct for the application's mechanical and thermal requirements. The reason it did so is that the box body design, as supplied to us, had no internal ribs on the inside face to resist the deformation the material's shrinkage anisotropy was always going to produce. It was a geometry problem, not a molding-process problem.
At Design-for-Manufacturability (DFM) review, we identified the shrinkage-anisotropy risk. This is the standard concern our engineering team is trained to flag for glass-filled semi-crystalline plastics on any shallow-flange geometry, and we did flag it. The problem was not that we missed the issue. The problem was what we did about it.
The correct response — and this is the lesson that has changed how we run DFM at Abery ever since 2011 — would have been:
"We are not going to cut steel on this box body until the internal geometry is revised to include structural ribs on the inside face, at adequate spacing, sized per rib-design guidelines (rib base thickness ≤ 60% of nominal wall, height ≥ 2.5× wall thickness, adequate draft). Nylon 33% GF plus this open-flange geometry will warp in production regardless of process parameters. Please issue a revised part CAD and we will re-quote."
That is not what we did. What we did was:
Flag the risk in writing to the direct customer's engineering team.
Accept their answer that the internal geometry could not be revised at that stage of the program because the ECU internals were already laid out around the existing envelope.
Propose that we would use assembly fixtures ("jigs") during sample inspection to hold the box body in the correct geometry for measurement, so that dimensional sign-off could proceed.
Cut steel and sample the mold.
Fixture the samples, present them for end-customer inspection, and receive sign-off approval.
Move into production of 5,000 pieces without the fixtures.
Reading step 3 back now, I want to be honest about what it actually was. It was a sample-approval workaround that let all three parties in the chain — us, the Tier-1, and JCB's engineering team — move a project forward that should not have moved forward until the box body was redesigned. The fixtures did not fix the underlying warpage; they hid it long enough for the samples to look compliant on an inspection bench. Every party in the chain preferred, in that moment, to close the design phase and get to production over redoing the CAD.
I want to be careful with the word "hid." Everyone involved knew the samples were being inspected with fixtures — this was not a secret. But the collective agreement in the chain was "if the fixtured samples measure correctly, we can proceed." That is a very different standard than "the parts, as molded in unattended mass production, will measure correctly." The first standard is a contractual convenience. The second is what actually matters when 5,000 parts land in a UK assembly line.
We had the technical authority to refuse the first standard. We did not use it.
The samples were signed off. We ran 5,000 parts. Without the sample-approval fixtures, the flange plane warped exactly as glass-fiber anisotropy predicted. The 5,000 parts arrived at the end customer, were inspected on a production assembly line — not on a fixtured inspection bench — and were rejected.
Our direct customer contacted us. They asked us to remake 5,000 parts and air-freight them by DHL express. The commercial position, on paper, was clear on both sides:
Their view: the parts are non-conforming in the field, we need replacement parts, please make them.
Our contractually clean position: the samples were signed off. That sign-off released our obligation on the specific dimensional characteristic that had been signed off. We were, strictly speaking, within our rights to invoice for a second run.
I want to explain why we did not take the contractually clean position, because it matters for how a serious buyer should read this case study.
One — the sign-off was jointly compromised, not one-sided. The fixtured-sample approach had been our proposal. We had made the workaround available. When it failed in production, the responsibility for that failure was not evenly distributed but it was also not zero on our side. Treating the sign-off as a full contractual release, when we knew the sign-off had been made possible by a fixture we had designed and applied, would have been legalistic in a way that damages relationships.
Two — the direct customer was in an impossible spot. They were between us and JCB. If we invoiced for the remake, they would either eat the entire cost themselves or have to renegotiate with JCB — a conversation that no Tier-1 wants to have on a program that is already late. Taking the loss on our side kept the failure contained inside the Chinese supplier's cost line, which was where a serious engineering-honest analysis said it belonged.
Three — we had built the mold, we knew the design limits, and we had accepted the workaround. Owning the outcome of that chain was the professionally honest thing to do.
So we remade 5,000 parts. We shipped them by DHL air freight to the UK. The direct customer covered the freight; we covered the material, the machine time, the labor, and the shortfall on the original tooling amortization. The net loss to Abery Mold on the project was approximately USD 10,000.
Every case study on the internet from a mold maker ends with "we learned a valuable lesson." Ours actually did, and I can point to the specific procedural change.
Since 2011, Abery Mold's internal DFM process for any glass-filled semi-crystalline material (glass-filled nylon, glass-filled PBT above certain fill fractions, glass-filled PPS, etc.) on any shallow, open, or flange-heavy geometry follows this rule:
Sample inspection may not use assembly fixtures to correct for anisotropic-shrinkage warpage. If the free-state part does not meet dimensional requirements, the mold or the part design is revised before sample sign-off.
Concretely:
Our DFM report on any glass-filled-nylon box, cover, or enclosure geometry now includes a warpage prediction — either via Moldflow / Moldex3D simulation for programs that justify the cost, or by our engineering team's structured checklist for programs that do not. The warpage prediction is shared with the customer before steel is cut.
If the predicted warpage exceeds the tolerance envelope, we will not quote sample-fixture inspection as a path forward. We will quote a design revision, typically internal rib addition, and if the customer cannot accept the revision we will decline the project rather than accept it on terms we know will fail.
Our chief engineer has written authority to refuse fixture-assisted sign-off on any mold in the categories above, and that authority does not require CEO sign-off to exercise. It is faster to refuse than to explain.
This rule has cost us at least three projects since 2011 that we would otherwise have accepted. We think it is worth it. The USD 10,000 we lost in 2011 has been the cheapest engineering-education investment Abery Mold ever made, and I would rather explain the rule honestly on our website than repeat the failure quietly on another program.
A few months after the remake shipped, the same Tier-1 customer placed a substantially larger order with us. Over the years since 2011, they have continued to work with us on new mold programs.
I want to be very careful about how I frame this. The larger order that followed is not the lesson of this case study, and I do not want it to be read that way. If I present the sequence as "we ate USD 10,000, the customer rewarded us with a bigger order, therefore eating losses is a good business strategy" — I have turned a real engineering lesson into a transactional hack, and I have taught the wrong thing to any junior engineer or salesperson at Abery who reads this article.
The lesson of this case study is in Section 6, and it is a procedural change in how we approve glass-filled-nylon molds. The outcome in Section 7 was pleasant but incidental, and it should not be used as an argument for accepting losses on future projects. The argument for accepting the loss on this project was that we had contributed materially to the failure through the fixtured-sample workaround, and taking the cost was the honest resolution of that specific project — regardless of what any future order might or might not have looked like.
A few things this project would tell me to check, if I were the buyer sitting on your side of the RFQ:
Does the supplier's DFM report include a specific warpage assessment for glass-filled semi-crystalline materials on any shallow-flange geometry? If it does not, they are quoting steel and tonnage without engaging with the actual failure mode of your application. Ask for the report before you place the PO.
Does the supplier propose fixture-assisted sample inspection as a way to hit sign-off? If they do, understand what you are agreeing to. Fixtured samples are a contractual convenience. Production parts are what land in your assembly line. Do not confuse the two.
Does the supplier have written authority for their engineering team to decline a project during DFM? This sounds procedural, but it is the single strongest indicator that the supplier is capable of saying no to bad geometry — which is what protects you from the failure mode described in this article.
On overmolded connectors in engineering thermoplastics (PBT-GF, PPS-GF, PA-GF over Molex, TE, JAE, or similar), does the supplier discuss preheating, insert-hold force, and post-mold anneal specifically for your material selection? These are the details that distinguish a real overmold shop from a general mold shop that does overmolding when asked.
If you want to have that conversation for a JCB, Caterpillar, Volvo CE, Komatsu, Hitachi, Doosan, or other off-highway ECU or telematics housing program, my email is at the bottom of this page. I answer serious construction-equipment RFQs personally.
This case study describes a project completed in 2011 in which the end application was destined for JCB via a UK Tier-1 electronics supplier. Our current relationship with the direct Tier-1 customer is protected under a mutual NDA; we do not claim direct approved-supplier status with JCB and do not represent ourselves as such. Buyers evaluating us for construction-equipment electronics housings should reference our capabilities on their own merit.
Because the specific engineering lesson — how not to use sample fixtures to release a mold on glass-filled-nylon warpage-prone geometry — is a lesson that matters to any buyer placing a similar program. Publishing successful projects only tells buyers what we can do; publishing this one tells them how we think when we made a real mistake, and what we changed as a result.
We will decline to quote the tool as designed. We will provide DFM feedback identifying the specific rib additions required, along with a warpage estimate for the current design and for the revised design. If the customer cannot revise the CAD, we will explain that we do not accept fixture-assisted sample sign-off on this material-geometry combination, and we will recommend they source the tool elsewhere. It has happened three times since 2011.
Yes. Overmolding of PBT-GF, PPS-GF, and glass-filled nylon around Molex, TE Connectivity, JAE, and similar industrial connector housings is an ongoing category for us. See our Overmold Injection Mold service page.
For a moderate-complexity two-tool family — one overmolded connector face plus one structural body — our standard lead time from PO to T1 samples is 50–60 days, including DFM, mold flow analysis on the glass-filled part, steel cutting, and initial trials. Add 10–15 days for T2 and final sign-off. For automotive-grade or off-highway PPAP-equivalent documentation, add an additional 15–20 days.
Tonney Shao is the founder and CEO of Shenzhen Abery Mold & Plastics Co., Ltd., an ISO 9001-certified injection mold maker and plastic parts manufacturer serving automotive, consumer electronics, medical device, industrial equipment, and consumer goods customers worldwide. Abery has built more than 3,000 molds since 2009 and operates a tooling and molding facility in Shenzhen, China. Tonney personally reviews every case study published on this site, and every DFM refusal decision above the threshold defined in Section 6.
Contact: tonney@a-mold.com · +86-13925214356
Case study published 6 August 2026. The direct customer's identity is withheld under a mutual non-disclosure agreement. The end-application manufacturer, JCB, is named in this case study to identify the technical context of the ECU housing family; Abery Mold does not claim direct approved-supplier status with JCB. Molex is a registered trademark of Molex, LLC. Material trade names PBT, Nylon (PA), and their glass-filled variants refer to generic material classes and are not attributed to any specific compound supplier's grade. The image on this page is a generic CAD render of the box body geometry family and is used to illustrate the technical points of this case study.
