My opinion is simple: when you choose an IoT module by unit price, you're making the most expensive decision of the project. That sounds like a paradox, but I've seen it play out more times than I can count. I'm not a salesperson. I'm a hardware sourcing specialist at an IoT product development firm. In the last seven years, I've coordinated more than 60 rush orders for prototype modules and connectivity components. This isn't a theory. It's a field report.
Let's start with the module everyone asks me about first. The Quectel EC25 LTE category is Cat 4, which puts it in the mid-speed tier—built for applications like in-vehicle gateways, high-bandwidth telemetry, and remote terminals. According to publicly listed Quectel specs, the EC25 supports downlink rates in the 150 Mbps range. That's a strong capability. But it's only the right capability if your product actually needs it.
Compare that with the Quectel BG95. The BG95 is a different tool: it's a low-power module covering LTE Cat M1 and NB-IoT, designed for battery-powered devices that send small amounts of data and then sleep. If you're building a water meter or a soil sensor, the EC25 is overkill. If you're building a vehicle tracker that needs live positioning and regular OTA updates, the BG95 might be underkill. Neither module is "better." The better module is the one whose total ownership cost fits your power budget, data requirements, and deployment environment.
What "LTE category" actually costs you
The "category" in a module spec isn't just a speed label. It determines the power amplifier design, the thermal dissipation, the antenna requirements, and the certification path. A Cat 4 module demands more from your power system than a Cat M1 module. That extra demand shows up as a bigger battery, a wider PCB trace, a more expensive voltage regulator, or a field failure that only happens at the worst possible moment.
Voltage drop: the silent TCO killer
One of the most expensive lessons I've learned came from a $1.20-per-unit difference. We approved a module that looked identical to the Quectel BG95 variant our hardware lead wanted. The specs were close, except the minimum supply voltage was slightly higher. We thought it wouldn't matter. It did. I don't remember the exact voltage levels now—don't hold me to a specific number—but the diagnosis was clear. During cold-weather field tests, the devices would boot loop when the battery was below 60%.
We fixed it by adding a boost converter and reworking the board. That cost us—or rather, it cost the client—about $4,000 in replacement parts and engineering time. The unit price saving vanished.
That's why I now run a voltage drop calculator before every design review. A module's datasheet says "3.0 to 4.3 V VCC," but that's not what the module actually sees. It sees the voltage after the battery connector, after the charging circuit, after a protection diode, and after a trace that's carrying a 2 A burst. At high current, a worn connector can drop 150 mV or more. A voltage drop calculator is a two-minute check that can save you a three-week diagnostic cycle.
Rugged devices and the DuraXV Extreme lesson
A client came to me with a rugged LTE handheld that reminded me of the Kyocera DuraXV Extreme. It was the kind of device that gets dropped on concrete, used in the rain, and left in a truck on a hot day. The mechanical design was solid. The electronics weren't.
The team had chosen a tiny module to save board space. It fit, but it required an external ultracapacitor to survive transmission bursts with the supplied battery pack. The ultracapacitor cost $1.90, the layout complexity added two weeks, and the module vendor's support was basically an email ticket that came back seven days later. The module itself was cheap. The system that made it work was not.
Rugged products like the DuraXV Extreme don't need the cheapest module. They need a module with predictable supply voltage, stable firmware behavior, and a vendor that picks up the phone when a field failure report comes in. That's a TCO decision, not a sticker-price decision.
The supply chain question: where are TVs made?
One question I hear constantly from procurement teams is "where are TVs made?" It's usually a supply-chain version of "how much risk am I carrying?" In electronics, the sharper version is "where are TVS made?"—the transient-voltage suppressors that protect your module's power input. If your component vendor can't tell you where those are made, or doesn't have a secondary source, that's a hidden cost.
The same logic applies to the module itself. A module might be designed in one place and manufactured in several others. That's not automatically good or bad. But if a supplier can't explain its manufacturing footprint, you're buying a surprise. I'd rather pay a slightly higher unit price for a vendor with transparent supply-chain answers.
Time is a line item
In March 2024, a client called me at 3 p.m. on a Thursday. They had a compliance test on Saturday morning and their modules—bought from a discount distributor—were trapped in customs because of a paperwork error. Normal replacement lead time was two weeks. With 36 hours to go, we sourced the same modules from an authorized distributor, paid $800 in expedited shipping, and had them delivered in time. The test slot was worth $30,000. Missing it would have delayed the product launch by at least a quarter.
I have mixed feelings about rush fees. On one hand, they feel like a punishment for poor planning. On the other, after you watch a fulfillment center split a small order into three shipments and move it to the front of the queue, you understand the cost. Either way, time is a line item. If you ignore it, your TCO calculation is fiction.
The TCO checklist for Quectel modules
When I'm triaging a rush order, I run through a short checklist. It works for the Quectel EC25, the Quectel BG95, and just about any other module.
- Unit cost at real quantity. Not a sample price. Not a 10k price you'll never reach. The number you'll actually pay.
- Integration cost. Schematic effort, extra regulators, antenna matching, EMI filtering.
- Certification cost. Does the module already carry the approvals your product needs? If not, can you afford the delay?
- Power cost. Battery size, charging time, voltage drop margin, low-temperature behavior.
- Time cost. Lead time, engineering hours, and the probability of a re-spin.
- Risk cost. Supplier stability, field failure rate, and how long a replacement would take if a batch is bad.
The lowest quote often loses on at least one of these. The highest quote rarely has to justify itself.
To be fair, price still matters
I'm not saying budgets are imaginary. They're not. If a $10 module gets your product to market and a $30 module doesn't, the cheaper one is the rational TCO choice—assuming it doesn't cause a field failure or a certification delay. The point is to calculate, not assume. "It looks fine" is not a cost analysis.
Granted, this kind of thinking takes more time at the beginning. You have to read datasheets, model power budgets, and interrogate suppliers. But that upfront time is exactly the time that makes rush orders unnecessary.
Bottom line
In my opinion, the only honest way to choose a module in 2025 is total cost of ownership. The Quectel EC25 LTE category is appropriate for mid-bandwidth products. The Quectel BG95 is appropriate for low-power connected sensors. Neither is "the best" outside a specific system. The best module is the one with the lowest total cost for your use case, your environment, and your deadline.
My experience comes from about 60 prototype and low-volume production orders. If you're buying a million units a quarter, your economics will be different. But the principle holds: unit price is not TCO. And in my work, TCO is the only number that matters.
