I handle component sourcing and failure analysis for a small contract electronics manufacturer. I’ve been in this role since 2018, and I’ve personally made—and documented—fourteen connector sourcing mistakes. Together they cost roughly $34,000 in scrapped parts, rework, and expedite fees. That number is embarrassing, which is exactly why the checklist below exists.
It focuses on something most people treat as a commodity: the 6-pin Molex connector between a power supply and the rest of the product. If your company builds, repairs, or specifies networking equipment, industrial controls, or anything else with an internal PSU, this is the checklist I wish I had on day one.
The failure that pushed me to write it down
Last September, one of our customers called about enterprise access points rebooting at night. Every reboot dropped clients for thirty to sixty seconds. Their IT lead had already run a “what is on my wifi” check twice: no unknown clients, no interference, no changes to the network. They assumed the AP firmware was the problem.
It wasn’t. Our field tech—who still carries a Kyocera Duraxv Extreme because it’s the one device that has survived every site visit—confirmed the RF side was clean. The real fault was inside the power supply assembly: the 6-pin Molex connector that carries 12 V from the PSU to the main board had an intermittent contact. Terminal retention had degraded over hundreds of thermal cycles. The connector looked fully seated, but a single wire could be made to lose contact with a light tug.
A replacement harness with the same six-pin connector resolved the issue. But why did the original fail at all? Because a “compatible” six-position connector had been substituted during a parts shortage. Similar shape, different material, different reliability. Nobody caught it because nobody had a checklist that said check.
The checklist I run now
I wrote the first version in March 2024, after the third connector-related rejection that quarter. Here’s the version my team currently works from.
1. Identify the connector family before you talk price.
Every week, an email shows up requesting “Molex connectors 6 pin” for a power supply. That phrase alone isn’t enough to order anything. It could be a Mini-Fit Jr. (4.20 mm pitch, common in power distribution), a smaller Micro-Fit 3.0, or an older .093-inch pin-and-socket connector that people call Molex out of habit. They have different terminals, different current ratings, and different locking mechanisms. If the supplier can’t send you a manufacturer drawing number, stop there.
2. Draw the pinout, then check it against the housing orientation.
My classic mistake happened in 2021. I approved a 6-pin pigtail for a 12 V distribution board without checking the wire-side view against the PCB header. Pins 4 and 5 were crossed. The order was 250 pieces, and it took a pull test on three samples to catch what the drawing should have caught in three minutes. Cost: about $1,400 in rework, plus a week of delay that I had to explain to a customer.
Get the drawing, flip the connector to the wire-entry side, and trace each cavity to its label. Every time.
3. Look at the terminal rating at your actual temperature.
Most buyers focus on pin count and unit price—and completely miss the terminal’s current rating under real conditions. If a datasheet says a terminal is good for 9 A, that number usually reflects a single circuit in free air at room temperature. Put six circuits in one housing inside a sealed power supply at 50°C ambient, and the safe current can drop by 30 percent or more. That derating is in the Molex published data, but you have to go looking for it.
Do the math before you pick the terminal, not after the first thermal test fails. A 6-pin Molex connector isn’t “six times the current”—it’s six times the heat.
4. Match the plating and housing to the environment.
The same connector family can ship with tin, gold, or selectively plated terminals, in housings meant for standard or high-temperature processing. Tin is fine for many indoor power supply applications. Gold is worth the cost when the product sees frequent mating cycles, high humidity, or corrosive air. If your product lives in a cabinet above a factory floor or near a coastline, do not pick the cheapest plating from a distributor dropdown. I’ve seen the aftermath; it is not pretty.
5. You don’t get to say “crimped fine” without measuring it.
This is the step most people skip after the first prototype. The terminal looks crimped. The wire doesn’t pull out when you tug it by hand. That is not a crimp test.
Per IPC/WHMA-A-620, the industry standard for wire-harness assemblies, acceptable crimps are defined by crimp height, profile, and pull-force limits. A proper pull tester costs less than a single failed field service call. We bought one in 2023 after a batch of hand-crimped samples passed visual inspection but failed vibration testing. We now pull-test five samples from every new reel and record the values. That decision has caught bad crimps three times in the last eighteen months.
6. Test the final assembly the way it will actually fail.
If the product is an access point, don’t test the connector on the bench at room temperature. Power the PSU through the connector, put the assembly in a thermal cycle profile that matches the install environment, and wiggle the harness the way a technician would. Intermittent opens often appear only when a housing is warm and a wire is flexed. That’s exactly how the failing connector in our September case finally revealed itself.
What I still see people get wrong
Three patterns keep showing up.
First, engineers treat the connector as an afterthought and let procurement buy from a photo. That’s not procurement’s fault if the drawing isn’t attached to the RFQ.
Second, teams confuse “mates with” with “manufactured to the same specification.” A six-position connector from another source can mate with a Molex header while still having weaker terminal retention or a lower temperature rating. A connector is a mechanical system, not just an electrical one.
Third, companies do a thorough design review, then forget to verify the parts that arrive. Six months later, the factory substitutes a connector to meet a ship date, and the failure mode comes back. The sole fix is a receiving check that includes the part number, the date code, and a pull test sample from the reel.
I would rather spend ten minutes explaining these checks to a customer than spend another week explaining why a connector failure took down their whole network. An informed customer asks better questions and makes faster decisions.
The first version of this checklist cost me $34,000 to learn. The current version has caught four bad production lots before they shipped. Copy it, adapt it, and if a supplier pushes back on any of these steps, treat that as a signal.
