Walk into any IoT planning meeting and the conversation is about sensors, dashboards, alerts, and which color the status badge should be. The cellular link gets one bullet point: “LTE backhaul, dual SIM.”

Then the deployment goes live and the cellular link is what fails first, fails hardest, and is the most expensive thing to fix once the gateways are mounted.

Here’s what we’ve learned about getting it right.

The carrier you tested with at the office isn’t the carrier at the yard

We have a desk drawer full of dev kits with AT&T SIMs in them. We’ve also been to enough customer sites to know that AT&T’s coverage map and AT&T’s actual signal at a yard 30 miles outside a metro are not the same document.

Before we ship, we send a tech with a multi-carrier signal meter to the proposed mounting location — not the site entrance, not the office trailer, the actual mounting location — and we measure RSRP, RSRQ, and SINR for every carrier we have a SIM for. The site survey is two hours. The cost of skipping it is a return trip with the right SIM and a customer who’s lost confidence in the install.

Antenna placement matters more than antenna spec

Default to a 4dBi panel antenna mounted as high on the building shell as you can reach. Higher beats fancier. A 9dBi log-periodic antenna pointed at the wrong tower performs worse than a cheap omni mounted six feet higher.

If you can see the tower, you don’t need much antenna. If you can’t see the tower, no antenna will save you — you need a different mounting location.

Dual-SIM is not optional. The failover logic is the question.

Every gateway we ship is dual-SIM, AT&T + Verizon by default, with T-Mobile available where it makes sense. The question isn’t whether to run dual-SIM — it’s how the failover behaves.

Naïve failover swaps SIMs the moment the primary loses signal. That’s wrong. Brief LTE drops are normal; if you swap SIMs every time, the secondary carrier’s session establishment overhead means more downtime, not less. We hold the primary for 60 seconds of confirmed loss before switching, and we don’t switch back automatically — we let the next scheduled health check decide.

Latency is the metric, not “bars”

A gateway showing four bars and 2-second round-trip latency is broken. A gateway showing two bars and 80ms round-trip is fine. We instrument every gateway with end-to-end ping latency to our broker, and that’s the number on the health dashboard. Signal strength is a leading indicator at best.

What we ship

The standard backhaul kit on an Echolo gateway install is:

  • Industrial cellular router, dual SIM, with PoE for the LoRa gateway and PoE cameras on the same pole.
  • 4dBi panel antenna mounted to the building shell.
  • Lightning arrestor inline on the antenna feed. Skip this once and you’ll learn why.
  • Battery backup sized for four hours of broker connectivity.
  • A health-check job that pings the broker every 30 seconds and reports RTT to Fleet.

Five components. Each one boring. The whole package is what makes the rest of the deployment possible.

The takeaway

The cellular link is the foundation. Spend the time on it. The dashboard the customer sees every day is built on top of a stack that fails first at the radio, and no amount of frontend polish makes a dropped connection legible.

Boring infrastructure carries the load.