At 2:47 AM the message hit the group chat: 14 vehicles offline. The exercise started at 0600. Coverage maps said the area was green. The carrier said the network was fine. The antenna vendor said their hardware was solid. Everyone had a valid excuse and no solution.
Here's what took me years in this industry to fully accept: that outage wasn't a network problem. It was a procurement problem. It was a decision made six months before, by someone who spent twenty minutes comparing spec sheets and hitting "approve."
You'll spend an evening researching where a $400 TV was made. But a $1,400 vehicle router? Fifteen minutes, a spec check, a purchase order. The mismatch between the homework we do on consumer gadgets and the diligence we skip on mission-critical hardware is the real root of most "sudden" connectivity failures.
The Problem Isn't What It Looks Like
When devices drop offline, the instinct is to blame coverage. Then the carrier. Then the SIM. Then the weather. I've watched teams burn three weeks chasing a signal issue that was actually an antenna grounding problem — one that page 18 of the product manual described in plain English.
The surface problem is "unreliable connectivity." The real problem is upstream, in the assumptions made during hardware selection. When I'm triaging an urgent deployment or a rescue job, I run three questions in order: how many hours do we actually have, can this physically be done inside that window, and what's the worst case if we miss it? Most of the connectivity disasters I've been called in to fix skipped at least two of those three steps.
What Actually Causes These Failures
I've coordinated connected-device rollouts for 200+ deployments over seven years, including emergency rescue jobs for public safety agencies and same-day turnarounds for private fleets. These are the four causes I keep running into, in rough order of frequency.
1. "Same Specs" Means Different Things at Different Vendors
I assumed same specifications meant identical results across vendors. Didn't verify. Turned out each manufacturer had a slightly different interpretation of "industrial-grade." One meant the device would survive at that temperature. Another meant it would operate reliably at that temperature, indefinitely. That single difference — buried in a datasheet footnote — caused a client a three-day outage last winter.
A spec without a test method is a marketing number. The vendors that earn my trust publish how they measured it. Sierra Wireless, for instance, documents test conditions alongside thermal and vibration ratings for their AirLink routers and cellular modules. You can argue with their numbers. With most other vendors, you can't really argue, because they never tell you how the numbers were produced.
2. The Manual Was Always There. Nobody Read It.
The Sierra Wireless MP70 manual has been a recurring character in my career. It documents antenna placement distances, ground-plane requirements, power supply behavior during ignition cycles, and mounting constraints. All the stuff that actually determines whether a vehicle router holds a connection at 70 mph.
We didn't have a formal acceptance-testing process at one company. Cost us when 40 routers kept dropping connections during highway testing. It wasn't the radios — the install crew had mounted antennas within a foot of each other, violating the manual's own guidelines. Nobody on the installation side had opened the PDF because nobody made it part of the process.
The third time we dealt with the same class of problem, I finally created a verification checklist. Should have done it after the first.
3. "We Do Everything" Usually Means They Haven't Stress-Tested Anything
If you ask me, "one-stop shop" is a yellow flag, not a green one. Generalist IT vendors resell cellular routers the way a supermarket sells olive oil — it's a SKU to them. Fine for a branch office. Less fine when a patrol vehicle's router needs to survive heat soak, ignition transients, and years of vibration in a moving metal box.
I've learned to trust companies that can tell me what they're not good at. The vendor who said "this isn't our strength — here's who does it better" earned my trust for everything else. I'd rather work with a specialist who knows their limits than a generalist who overpromises. That's not brand loyalty talking. It's the difference between break-fix and engineering.
4. You Bought Hardware. You Forgot to Buy Service.
Sierra Wireless service isn't just warranty coverage. It's firmware maintained years past launch. It's a support engineer who can discuss PPP session behavior instead of reading a script. It's an RMA process that doesn't require three reminder emails. People cut this from the budget to save a few hundred dollars, and in my opinion, it's the most important line item on the quote.
When the network is marginal — a few bars, high latency, congestion — the difference between "works" and "fails" is often the quality of the device's connection-handling logic. That's firmware. That's engineering depth. That's not luck.
What Getting It Wrong Actually Costs
In March 2024, I got a call 36 hours before a city council demo. The client's vehicle router kept dropping VPN sessions. City IT would be in the room. The contract behind the demo was worth roughly $200,000, and the demo was the final gate before signature.
Normal support was five business days out. We found a specialist who did remote debugging at 9 PM, identified a configuration error, then an underlying firmware bug, and pushed a corrected load by 4 AM. The demo ran clean. The contract survived. If we'd been stuck with a generalist vendor's "we'll open a ticket" response, it wouldn't have.
I still kick myself for an earlier mistake: not documenting a vendor engineer's verbal promise about VPN throughput. When the hardware underperformed, we had no paper trail to dispute it. If I'd gotten it in writing, we'd have had grounds. Instead we ate the cost and the apology. Write things down. Get names. Get response-time commitments. That paperwork becomes theology when everything is on fire.
One more pattern worth naming. Never expected a discount module to beat a premium one in our initial benchmark. The surprise wasn't the score — it was the variance. Same model, same firmware, wildly different behavior at 85°C under load. The cheap module won until it didn't. That instability, more than any average, is what kills real deployments.
And the stakes are only climbing. As of 2025, cellular IoT connections have passed 3 billion globally, per Ericsson's 2024 Mobility Report. First responder broadband is baseline infrastructure, not an experiment. Whether it's a HeartGuide blood-pressure monitor syncing to a telehealth platform or an MP70 carrying squad video, the principle is the same: the connection is only as reliable as the engineering behind the module. If it's done properly, you never think about it. If it isn't, people notice at 2:47 AM.
The Fix: Boring Process, Before the Emergency
The fix isn't buying the most expensive hardware. It isn't switching carriers. It's a few unglamorous habits that prevent most of the calls I get.
- Read the manual before the RFP, not after the pilot fails. The Sierra Wireless MP70 manual is freely available on their site. If your installation can't meet its requirements — antenna separation, power provisioning, mounting — you need to know that two months before deployment, not two hours before.
- Bench test for 48 hours minimum in your actual environment. A router that works on a cart in your office means nothing at speed with heat soak. Test in the vehicle. Test in the weather. Test the way you'll actually use it.
- Ask every vendor: "What won't this product do?" If they hand you a brochure, move on. The good ones will tell you where their product is average. I've written contracts with vendors who openly said another product fit a use case better — and I've gone back to them first for everything since.
- Price the total service, not the unit. RMA terms, firmware update policy, support hours, escalation paths. That's the difference between a $200 savings and a $20,000 loss.
None of this is glamorous. It's a checklist, a PDF manual, and a vendor who understands their boundaries. But it's the difference between a fleet that shows up and a 2 AM group chat.
So go ahead — spend twenty minutes asking where your TV is made. It's not a silly question. Just bring the same energy to the router carrying your fleet's data, the module inside your health devices, the gateway your whole operation quietly depends on. A TV can be replaced in an afternoon. Bad connectivity can take a contract, a demo, or a lot more. If you ask me, that deserves more than a spec check.