Writing, when there's something worth saying.
Technical field notes, values pieces, and the occasional post-mortem. No SEO filler. No "5 trends in IoT for 2026."
- 4 min readproduct
The machine that left at 2am
Presence and last-seen won't paint a live dot on a map — and shouldn't pretend to. But the moment a tag stops being heard, and the last hub that heard it, is a record worth having. Here's what that boundary honestly catches, and what it doesn't.
Read → - 5 min readproduct
You don't have a 40-machine fleet
The asset list says forty machines. The run-hours say twenty-eight are earning and twelve are insurance you're paying to store. Utilization doesn't just tune your maintenance — it decides what you buy next.
Read → - 5 min readmobile
Building Fleet for the field
The office runs Fleet in a browser tab. The field runs on a phone — in the sun, in gloves, in a dead zone — and that's a different app. Notes from building the mobile half of the platform, and the Bluetooth trick that turns every phone into a roaming hub.
Read → - 4 min readproduct
Location isn't utilization
A map full of dots tells you where your equipment is. It tells you nothing about whether any of it is working. Those are two different questions, and most fleets only own hardware that answers the first.
Read → - 5 min readtechnical
Getting your data out
Somewhere in every buying cycle a careful customer asks the question that separates a tool from a trap — "can I get my data out?" A dashboard is where data goes to be looked at. The fleet data that matters has to live in the systems you already run. Here's how we think about the way out.
Read → - 5 min readproduct
Idling isn't working
At month-end the 50-ton excavator shows 190 hours — the busiest number on the yard. Then you find out 70 of them were the engine running while nobody touched a lever. Hours tell you the machine was on. Utilization tells you whether it earned its keep, and those are not the same number.
Read → - 5 min readfield
Tagging the whole yard
The PO is signed and a box of forty tags shows up. Between that box and a dashboard that means anything sits the least glamorous step of the whole deployment — giving every machine an identity the system can trust. Do it wrong once and you'll chase phantom assets for months. Here's how to do it right, and why it isn't the six-month project everyone fears.
Read → - 3 min readproduct
The 490-hour service
We wrote a while back that continuous CMH lets you schedule the 500-hour service at 490 instead of "when someone notices." That was half the story. Here's the other half — turning the number into a job that someone actually does.
Read → - 5 min readfield
The first 30 days of a deployment
Signing the PO is the start, not the finish. The first month decides whether a deployment becomes infrastructure the site forgets it has, or a support ticket that never closes. Here's the arc — a composite of the ones that went well.
Read → - 6 min readfield
Designing for the dead zone
We've written about picking a good cellular link. This is the other half of the problem — what the system does when there's no link at all, for a minute, an hour, or a whole shift. Off-highway, that isn't an edge case. It's Tuesday.
Read → - 5 min readhardware
What a BLE tag's battery actually lasts
The datasheet says five years. The tag on the excavator went dark in fourteen months. Neither number is a lie — they're answers to different questions. Here's what actually drains a coin cell in the field, and why tags die all at once instead of slowly.
Read → - 4 min readfield
Rental hours and the return-day argument
The rental house's invoice says the machine ran 180 hours. Your operator swears it was closer to 140. Both of you are reading a meter — only one of you was watching it the whole time. Here's how telemetry settles it without starting a second fight.
Read → - 5 min readproduct
A day in Fleet
The new Fleet App ships soon. Instead of a feature list, here's what a Tuesday looks like for an ops lead with it open — from the 7am inbox paragraph to the 5pm audit log.
Read → - 6 min readfield
Why off-highway equipment is its own IoT problem
A connected forklift is a known problem. A haul truck on a mine bench, an excavator on a remote easement, or a tractor halfway across a section — those are different problems wearing the same name.
Read → - 6 min readfield
Solar and battery sizing for remote sites: the math nobody does
Two panels and a deep-cycle battery is not a power system. It's a power system that works in May. Here's how we actually size off-grid IoT installs so they survive February.
Read → - 5 min readfield
Site surveys: the things you measure that aren't on the spec sheet
A spec sheet tells you what the hardware can do. A site survey tells you whether it will. Two hours on a roof saves a six-week return-trip story.
Read → - 4 min readtechnical
AI in IoT: what actually runs in production
Every IoT vendor has an AI story. Here's where machine learning genuinely earns its place in a real pipeline — and where it quietly gets replaced by a for loop.
Read → - 4 min readproduct
CMH and the warranty conversation
When a $40,000 engine fails at 1,840 hours and the warranty runs to 2,000, what counts as proof? The boring bookkeeping that decides whether the claim gets paid.
Read → - 3 min readproduct
CMH from telemetry vs CMH from the clipboard
The dashboard says 4,820 hours. The operator's logbook says 5,180. The gap is real, it's persistent, and it's costing someone money. Here's how to think about it.
Read → - 3 min readproduct
What "running" actually means
Every yard tracks "machine hours" and every yard means something slightly different by it. The definition you pick changes the number by 20% or more, and decides whether your service intervals work.
Read → - 3 min readfield
Cellular backhaul, the unglamorous part of IoT
Everyone wants to talk about sensors and dashboards. The thing that decides whether your deployment works is the cellular link, and almost nobody designs it carefully enough.
Read → - 2 min readvalues
Why we don't name customers on our marketing site
There are two stories in every B2B deployment — the one the vendor wants told, and the one the customer actually lived. We default to neither.
Read → - 3 min readfield
LoRaWAN in the wild: notes from a few real deployments
Datasheets talk about 10-mile range and decade-long battery life. Fish farms, remote sites, and metal-heavy yards tell a different story. What we actually see in the field.
Read → - 3 min readproduct
When predictive maintenance isn't magic
The industry loves to talk about AI + IoT for equipment management. Most of what actually ships in production is thresholds, trendlines, and someone willing to look at the dashboard.
Read → - 2 min readtechnical
MQTT at scale, field lessons
What we've actually learned running an MQTT pipeline in production — the mistakes we made, the instrumentation that saved us, and the parts of the protocol nobody warns you about.
Read → - 3 min readtechnical
Data pipelines without drama
Every IoT company eventually writes one. Every one of them is secretly held together by queues, dead-letter exchanges, and the hope that nothing fails at 3am. Here's what ours looks like, plainly.
Read → - 2 min readfield
What BLE tracking in a heavy-equipment yard actually looks like
A field note from a gateway install. Rust, forklifts, unexpected RF geometry, and what a 12km datasheet range means in a building made of steel.
Read →