We Bought Four Different Safety Products Wrong. None of It Was the Product's Fault.

The third week of September 2022, I sat at my desk with four order confirmations that had each either come back, been stalled, or been redone. One was a Tyco fire alarm upgrade for a warehouse extension. One was a hog wire fence for the back lot. One was a dog fence boundary between two of our units. One was a fall arrest system for the crews working on the elevated platform.

Four different safety products. Three reworked, one paused on site. My first reaction — and I want to be honest about this — was to assume we'd been sold bad product.

That assumption was wrong. Or rather, not wrong so much as convenient. Every single piece of hardware checked out on paper. The Tyco fire products we'd selected met spec. The hog wire fence was cut and glued correctly. The dog fence didn't have a defective component. The fall arrest system's harness was a standard, correctly-rated piece.

The problem wasn't the products. It was what I asked the products to do, and what I never verified before signing purchase orders.

For three years I assumed the brand name was the safety guarantee

When I first started procuring safety gear for our facilities, I believed the brand name was the safety guarantee. If we bought Tyco fire products, or a big name everybody recognizes from the electrical catalog, then safety was somebody else's problem already solved.

That assumption quietly cost us money in at least three of four categories that came back to me in September. Let me break the reasoning apart, because it took me until Q1 2024 to actually understand it.

Safety products are systems, not products

Take the easiest one to misunderstand. If you've ever asked what is a fall arrest system and gotten the answer "a harness," you've already run into the trap. It's not a harness. It's a chain: anchor point, connector, shock absorber, harness, and — the part almost everyone skips on the first draft — a rescue procedure.

OSHA 1926.502(d)(15) is unambiguous about this: fall arrest anchors have to support at least 5,000 pounds (22.2 kN) per employee, or be designed, installed, and used under supervision of a qualified person with a safety factor of at least two. The harness is the easy part of that entire requirement. The anchor is where the requirement actually lives.

We installed the 2022 system with a correctly-rated harness, clipped to an eyebolt on an older I-beam whose capacity nobody had actually tested. The rigger did it in a morning. The safety inspector found it in the afternoon and stopped the whole install. Two install bills, one harness that worked fine, one anchor that had to be redone. The product was right. The system wasn't.

Tyco fire alarms follow the same logic. The alarm unit itself is a piece of hardware. A Tyco fire alarm system is a loop, a control panel, the device locations, notification coverage, and — this is the part that bites procurement people — the monitoring on the loop itself. NFPA 72 governs the system, not the device. (Note to self: this is exactly the sentence I keep having to repeat to our new project managers.)

We compared specs instead of failure modes

Spec-sheet comparison is where the hog wire fence got away from us. The parameters were: 12.5 gauge, 48 inches high, 2x4 mesh, 100-foot rolls. Every supplier in front of us matched those numbers. I ordered on those numbers. Within a week, the dogs had gone over it.

Here's what I missed, and it's not a spec problem. I bought a confinement fence. What I actually needed was a jump-restriction fence. Those are different sentences, and no spec sheet knows the difference. The parameters were right for the failure mode I thought I was buying against. Wrong for the failure mode that actually happened.

The dog fence case had the same shape. "Dog fence" in the catalog can mean two completely different products: a physical perimeter barrier, or a buried-wire electronic boundary that requires training and equipment beyond the wire itself. Electronic boundaries behave very differently over long runs — you need surge protection roughly every 1,000 feet of boundary loop, and the entire approach depends on training, which is an operational variable, not a spec. I ordered the wire, never thought about the training, and the dog pushed through the correction zone by the end of week one. That little decision cost roughly $400 in trainer time plus a temporary physical fence. (Ugh, again.)

I knew I should've written down the failure mode before ordering — that's the whole point of doing this in writing — but thought "it's just a fence, what are the odds." The odds arrived with a very determined terrier mix and a hole under the gate.

The question we should have been asking

Before every one of those four orders, the question I was asking myself was: "is this product good?"

The question I should have been asking is: "how will this product fail, and will I be able to see it failing before it becomes a real-world event?"

Try it on the Tyco fire alarms. Asking "is this alarm certified" is a much weaker question than "if the loop runs open between this device and the panel, where is the first indication that anything is wrong." That second question — the supervisory signal — is the NFPA 72 territory that actually determines whether the alarm system is functioning. A product datasheet cannot answer it. Only the system design can.

Try it on the fall arrest system. The failure modes that matter are anchor failure, mis-connection, and swing-fall from an unplanned anchor angle. A harness that passes every mechanical test still fails on all three. The system, again.

What the wrong orders actually cost

Numbers I can stand behind, because they came out of our own purchase records between September 2022 and December 2023:

  • Fall arrest anchor rework (re-engineering, re-install, reinspection): about $1,900, and two days of a crew standing around.
  • Tyco fire loop correction and re-certification on the affected leg of the extension: two days of delay, roughly $2,400.
  • Dog fence remediation: ~$400 in training plus temporary fencing, plus one very confusing conversation with a neighbor.
  • Internal team time spent redoing the same category of error three times. I can't put a clean number on this one, and I've stopped pretending I can.

Add up the duplicated labor, the delays, and the two re-inspections, and the "cheap" version of those projects was never cheap. It was just invoiced in pieces.

The six-point pre-order check we've been running for 18 months

I built the checklist in Q1 2024 after the third rejection of the same fiscal quarter. It's not glamorous, but it's caught 47 potential problems before they became purchase orders in the last 18 months. Rough estimate of avoided rework: somewhere north of $8,000.

  1. Draw the system, not the SKU. Fall arrest: anchor, connector, absorber, harness, rescue. Tyco fire alarm: detection, panel, notification, supervision, monitoring. If any element can't be filled in, the order is not ready.
  2. Write the failure mode down. For every component, one sentence: how does this fail, and how do we find out before it matters? If you can't write that sentence, you don't understand the product yet.
  3. Specs describe, they don't decide. Numbers on a datasheet are a description of the product. They are not a statement about your use case.
  4. Cross-check with a second supplier. Same requirements, second quote, and one direct question: "what's the failure mode of the spec I just gave you?" The answer tells you more than the price does.
  5. Ask the installer, not the salesperson. Installers see the failure modes weekly. Salespeople see the datasheets. Whoever is going to be on-site should sign off before the PO goes out.
  6. Buy one unit, one zone, one run first. One roll of hog wire, one alarm loop, one dog boundary. Run it a month. Then scale.

A short close

None of this is really about the products, and I want to be clear I'm not saying the manufacturers are at fault. The Tyco fire products we specified were correct. The fence hardware was correct. The problem was that I was buying products when the site needed systems, and I was comparing spec sheets when the site needed failure modes.

Swap one question before your next order — "is this product good?" becomes "how will this product fail?" — and your purchase records get cleaner very quickly. At least, that's been my experience with four of the four categories above.

Five minutes of verification has consistently beaten five days of correction. That's the entire lesson, and I paid for it three times before I wrote it down.


Leave a Reply

Your email address will not be published. Required fields are marked.