Daniel sent us a question that starts with Euroboxes and ends somewhere in the physics of micro-deformation. Here's what he wrote.
"As we've discussed in countless episodes on this podcast, I have moved into the exciting world of industrial storage and Euroboxes. The shelving that the industrial shelving store sold is rated for two hundred kilograms per shelf. We have four Euroboxes stacked on it, and I don't know how much each one weighs. I'm guessing that we're not even close to the margin, but I was wondering: if one wanted to take away the guesswork, what might be the most efficient way of doing this?"
He then lays out the two approaches. One — "the purest way: actually measure the load spread out across the entire width of the shelf." Two — "a weight sensor underneath each Eurobox, and the total weight on the shelf is computed by simple addition." And then the practical ask: "I know that weight sensors exist, and my smart home is equipped with Zigbee, Matter, and Wi-Fi. Are there any that are actually practicable for this application?"
The thing about this question — it's really two questions wearing the same coat. There's the engineering question of how to measure weight on a shelf, and then there's the consumer-electronics question of whether anyone's packaged that into something you can just buy and pair with your phone. And the answers point in completely opposite directions.
One's a solved problem and the other barely exists.
The sensing technology is ancient and cheap and well-understood. The consumer product category is... let me put it this way. I went looking for Zigbee weight sensors in the standard device database. The category page for scales returns a four-oh-four.
A four-oh-four is a pretty definitive answer.
It's not that the page moved. It's that the category doesn't have enough devices to justify maintaining it. Zigbee weight sensors are so niche that the canonical open-source smart-home project hasn't bothered to populate the page. That tells you something about the market.
So the short answer to "are there any I can just buy" is no. But I feel like we should earn that no by explaining what's actually happening under the hood, because the DIY path is genuinely interesting and absurdly cheap.
Let's do that. The question is really about two different ways of knowing — measuring the shelf as a system, or measuring each box as a component. Let's unpack what each approach actually requires.
Before we can judge which approach is better, we need to understand the physics underneath both of them. How does a load cell actually work?
A load cell is deceptively simple. It's a block of metal — usually aluminum or steel — with a specific shape machined into it. Glued to that metal block is a strain gauge, which is basically a very fine wire arranged in a zigzag pattern. When you apply weight, the metal block deforms microscopically. The strain gauge stretches or compresses along with it, and that changes its electrical resistance. Measure that resistance change, and you've measured the weight.
Micro-deformation. The shelf bends by less than a human hair, and that's your signal.
That's the whole game. And the key component that makes this usable for a DIY project is something called the HX711. It's a twenty-four-bit analog-to-digital converter designed specifically for load cells. Adafruit sells a breakout board for about five dollars.
Twenty-four-bit. Why does that number matter?
Because the voltage changes we're talking about are tiny. A load cell typically outputs a few millivolts at full load. If you tried to read that with a standard ten-bit ADC — the kind built into most microcontrollers — you'd get basically noise. The twenty-four-bit resolution gives you over sixteen million discrete steps. That's what lets you detect a few grams of change on a surface that can hold two hundred kilograms.
So the resolution is what makes the whole thing possible. Without it, you're just guessing.
Right. And the HX711 also handles the amplification and filtering. It's a complete front-end for load cells. You wire the cell to the HX711, the HX711 talks to your microcontroller over a simple two-wire interface, and suddenly you've got weight readings in your code. Adafruit's tutorial on this is the canonical starting point — it walks you through wiring, calibration, and reading values with an Arduino or ESP32.
That's the per-box approach in a nutshell. A load cell under each Eurobox, an HX711 to read it, and a microcontroller to do the math. But Daniel also asked about the other approach — measuring the load across the entire shelf width. That's what warehouses do, supposedly.
Supposedly is the right word. The image people have is that every Amazon warehouse shelf is instrumented with weight sensors, tracking inventory in real time. The reality is that most warehouses still use manual weight checks or simple mechanical limits. Smart load-sensing shelving exists, but it's a premium feature, not the norm. And when it is deployed, it's expensive for reasons that have nothing to do with the sensing principle.
What makes it expensive?
The sensing principle is the same — strain gauges on metal elements. What changes is the scale and the integration. An industrial smart shelf uses multiple load cells distributed across the shelf structure, all wired to a central controller. The cells themselves are ruggedized for industrial environments — sealed against dust and moisture, rated for millions of load cycles, individually calibrated and certified. The controller handles temperature compensation, drift correction, and fault detection across the whole array. And the whole system needs to be installed, calibrated on-site, and maintained.
So it's not that the physics is different. It's that the packaging is built for a warehouse, not a garage.
And the calibration. That's the hidden cost of any weight-sensing project, and it's where the industrial approach and the DIY approach converge on the same pain point. Load cells drift. Temperature changes affect the readings. Every time you power-cycle the system, the zero point might shift slightly. To get accurate readings, you need a known reference weight — something you trust — and you need to run through a calibration routine that maps the raw ADC values to actual kilograms.
How bad is the drift in practice?
For the kind of load cells Daniel would be using — the fifty-kilogram bar-type cells that cost about ten dollars each — the drift is typically on the order of a few grams per day once the system has thermally stabilized. The bigger issue is the initial calibration. You need to place a known weight on the cell, record the reading, place a second known weight, record again, and then your software fits a linear calibration curve. If you skip this step, your readings are meaningless numbers, not kilograms.
So you need something heavy that you already know the weight of. What do people use?
A bag of flour from the kitchen. A dumbbell. A gallon of water, which is about three point seven eight kilograms. Anything where you trust the labeled weight. The Adafruit tutorial walks through this — you set up a known weight, run the calibration sketch, and it spits out a calibration factor you hard-code into your project.
That's the labor nobody talks about in the YouTube build videos. The thirty seconds of calibration footage is actually an hour of fiddling.
And if you're doing four separate sensors, you do it four times.
Let's talk about the wiring, because this is where the two approaches Daniel described actually diverge in practice. For the whole-shelf approach, you'd want multiple load cells combined into a single reading. For the per-box approach, you want four independent readings.
The whole-shelf approach uses something called a Wheatstone bridge configuration. You can wire multiple load cells together so their signals sum electrically before they even reach the ADC. Four load cells, one HX711, one combined weight reading. This is how every digital bathroom scale works — four cells in the corners, one reading. It's elegant, it's cheap, and it gives you the total weight with a single calibration.
But you lose per-box information.
You lose all spatial information. You know the shelf has a hundred and thirty kilograms on it, but you don't know if that's one box or spread across four. For Daniel's use case — four Euroboxes, wanting to know if he's approaching the two-hundred-kilogram limit — the total is what matters for safety. But the per-box data is actually useful for a different reason.
Knowing which box is the heavy one.
If one Eurobox is loaded with eighty kilograms of books and the others are light, you might want to redistribute. The shelf rating assumes reasonably even load distribution. Stacking all the weight on one side creates a bending moment that the rating doesn't account for.
So the per-box approach gives you more useful information, but it costs more in hardware and complexity. Four load cells, four HX711s — or you can multiplex them — four calibrations, four data streams to manage.
And the question Daniel asked about summing the weights — does the error accumulate to the point where the total is unreliable? Let's put numbers on that. A typical load cell accuracy is plus or minus zero point zero five percent of full scale. On a fifty-kilogram cell, that's plus or minus twenty-five grams. Four cells, worst case, the errors all point the same direction — that's a hundred grams of potential error on the total. Against a two-hundred-kilogram shelf rating, a hundred grams is less than the weight of the smartphone you'd use to read the data.
So the error accumulation argument is technically true and practically irrelevant.
At this scale, yes. If you were building a laboratory balance, you'd care. For a garage shelf, the error budget is smaller than the uncertainty in whether you remembered to include the weight of the Euroboxes themselves.
Which brings us to the commercial landscape. Daniel asked whether there are off-the-shelf weight sensors that work with Zigbee, Matter, or Wi-Fi. You already mentioned the four-oh-four.
Let me be thorough about this, because it's a interesting market gap. Zigbee has device categories for door sensors, motion sensors, temperature sensors, humidity sensors, light sensors, pressure sensors, vibration sensors, water leak sensors — the list is long. Weight sensors are conspicuously absent. There are a handful of Zigbee body-weight scales on the market, but they're closed devices that report through their own app, not generic weight sensors you can integrate into a home automation system.
And Matter?
Matter, as of mid-two-thousand-twenty-six, has no standard weight-sensor device type. It covers contact sensors, motion sensors, temperature, humidity, pressure, flow, light, occupancy — the usual environmental and security suspects. Weight sensing isn't in the spec. It might get added eventually, but the demand signal is weak. Most people don't need their shelves to report weight data.
Most people don't need their shelves to report weight data. That's the sentence that explains the entire market.
It really is. The use cases are niche: inventory tracking for a home workshop, monitoring consumables like pet food or flour, or — as in Daniel's case — safety verification for a shelving unit that's rated higher than you think you need. None of these are mass-market problems.
Wi-Fi scales exist, but they're body-weight scales. They expect a human to stand on them for ten seconds, they report to a health app, and they're not designed for continuous load monitoring or distributed weight sensing.
And they're not hackable in the way you'd need. You can't take the load cells out of a Withings scale and repurpose them for shelf monitoring — well, you could, but at that point you're doing the same DIY work with worse components.
So let's build the realistic DIY path. If Daniel wants to do this, what does the bill of materials look like?
For the per-box approach — which I think is the right call for his use case — you need four fifty-kilogram load cells. These are the bar-type cells, about thirteen dollars each from Adafruit, cheaper from the usual import sources. One HX711 per load cell, or you can use a four-channel multiplexer to share one HX711 across all four. Let's say four HX711s at five dollars each for simplicity. An ESP32 microcontroller, about eight dollars — that gives you Wi-Fi and Bluetooth, plenty of GPIO pins, and enough processing power to handle four sensor streams.
Plus wiring, a breadboard or perf board, maybe a project box to keep it tidy. Under sixty dollars total.
Under sixty dollars. And the software side is well-trodden ground. The HX711 has Arduino libraries that handle all the low-level communication. You write a sketch that reads each sensor, applies the calibration factor, and reports the values. From there, you can push the data to whatever smart home system you're running — MQTT over Wi-Fi is the standard approach, and Home Assistant will happily ingest MQTT sensor data.
MQTT over Wi-Fi, not Zigbee. So the sensor assembly talks to the network over Wi-Fi, even if the rest of the smart home is Zigbee.
That's the pragmatic path. You could add a Zigbee radio to the ESP32 — there are modules that support both — but at that point you're adding complexity for protocol purity. The ESP32 already has Wi-Fi. Use it. The data ends up in the same place.
And if Daniel doesn't want to build a four-channel sensor array and write firmware? What's the minimum viable version of this?
A fifteen-dollar digital luggage scale and five minutes of manual checking. Pick up each Eurobox, weigh it, write down the number, add them up. If the total is under a hundred and fifty kilograms — which gives you a comfortable margin below the two-hundred-kilogram rating — you're done. No soldering, no calibration, no MQTT topics.
This is the part of the episode where we admit that the smart-home integration is solving a problem Daniel may not actually have.
He said it himself — "I'm guessing that we're not even close to the margin." The guesswork is probably correct. The question is whether removing the guesswork is worth sixty dollars and an afternoon of tinkering, or whether a luggage scale and a notepad achieve the same outcome in five minutes.
But that's not the spirit of the show, is it? Daniel didn't ask "should I weigh my boxes." He asked "how would one build a system to weigh them continuously."
And in that spirit — the build is achievable. The Adafruit tutorial is the place to start. Get one HX711 and one load cell working first. Learn the calibration workflow. Once you've got one sensor reporting reliably, scaling to four is mostly a matter of copying the circuit and managing the additional data streams.
What about mounting? A load cell needs to be supported at both ends with the weight applied in the middle, or cantilevered depending on the type.
The bar-type cells are designed to be mounted at one end, with the load applied at the other. You'd build a small platform for each Eurobox — a piece of plywood with the load cell screwed to the bottom, and a foot or spacer at the other end so the cell is the only load path. The Eurobox sits on the platform, the platform sits on the shelf through the load cell. It's not elegant, but it works.
Industrial chic.
Garage-core. It's a whole aesthetic.
Let's talk about the safety margin, because this is where the two-hundred-kilogram rating gets more complicated than it sounds.
The rating on the shelf is a static load rating. It means the shelf can support two hundred kilograms evenly distributed, sitting still, under ideal conditions. Real-world conditions are never ideal. If you're sliding a heavy Eurobox onto the shelf, that's a dynamic load — the impact force can briefly exceed the static weight. If the weight is concentrated on one edge, you're creating a bending moment that stresses the shelf differently than the test conditions. And over time, materials fatigue. Particle board sags. Metal brackets develop micro-cracks. The practical safe limit is probably more like sixty to seventy percent of the rated capacity.
So a hundred and twenty to a hundred and forty kilograms, not two hundred.
That's the rule of thumb. And that's why the measurement question matters — not because Daniel is likely to hit two hundred kilograms, but because knowing the actual load tells you how much margin you have. If the total is eighty kilograms, you've got plenty of headroom. If it's a hundred and sixty, you might want to redistribute or lighten the load.
And that's information a luggage scale gives you once, but a sensor system gives you continuously. The value of continuous monitoring is catching changes — you add a box of tools, the weight jumps, and you know immediately whether you're still in the safe zone.
Or you set an alert threshold. Home Assistant can watch the MQTT topic and fire a notification if the total crosses, say, a hundred and fifty kilograms. That's the smart-home value proposition — not the number itself, but the automation around the number.
Now that we understand the mechanism, the real question is: what can you actually buy, and what do you have to build yourself?
The honest landscape in mid-two-thousand-twenty-six: there is no consumer smart shelf product. The Zigbee ecosystem doesn't have a weight sensor category worth naming. Matter doesn't have a device type for it. Wi-Fi scales exist but they're body-weight devices with closed ecosystems. If you want a shelf that reports its weight to your smart home, you are building it yourself.
And the building is surprisingly accessible. The HX711 has been the standard hobbyist load-cell interface for years. The libraries are mature. The tutorials are plentiful. The cost is low.
The one thing I'd caution about — and this is where I see DIY projects stall — is the mechanical mounting. Getting consistent readings requires that the load cell is the only load path. If the Eurobox is also touching the shelf directly, or if the wiring is pulling on the cell, your readings will be inconsistent in ways that are maddening to debug. Spend time on the mechanical design. The electronics are the easy part.
That's a sentence you don't hear often. The electronics are the easy part.
In this case, they are. The HX711 does all the analog heavy lifting. You're basically wiring up a digital sensor at that point. The hard part is making sure the physical world cooperates with your measurement setup.
So we've established that DIY is the only real path. Let's talk about what that build actually looks like and whether it's worth your Saturday.
The build breaks down into three stages. Stage one: get one load cell and one HX711 talking to an ESP32, run the calibration sketch with a known weight, and verify that you can read stable weight values. This is an evening's work if you've never touched a load cell before, an hour if you have.
Stage two?
Stage two: build the mechanical platforms for the Euroboxes and mount the load cells. This is woodworking, not electronics. Measure carefully, make sure each platform is level, and verify that the load cell deflects freely without binding. Test each platform individually with the calibrated electronics from stage one.
And stage three is the integration — four sensors reporting to Home Assistant, with a dashboard showing per-box weights and a total.
Optionally an alert if the total crosses your chosen threshold. The whole project is maybe two weekends if you're being careful, one solid Saturday if you're focused.
The per-box approach also gives you something the whole-shelf approach doesn't: you know which box is heaviest. For a garage storage system, that's actually useful. You might discover that the box you thought was light is actually forty kilograms, and the box you've been treating as heavy is mostly packing material.
The data often surprises you. I've done enough weight measurements in various contexts to know that human intuition about weight is remarkably unreliable. We're influenced by density, by shape, by how awkward something is to carry. A compact box of bolts feels heavier than a large box of blankets, even when the blankets weigh more.
The sensor system doesn't just remove guesswork — it corrects for systematic misperception.
Which is a fancy way of saying you'll be wrong about which box is heaviest until you measure it.
One last thought on where this category is heading — and what we'd love to hear from you.
The open question is whether Matter will eventually add a weight-sensor device type. The infrastructure is there — Matter already has a well-defined data model for sensors, and adding a weight cluster wouldn't be technically difficult. But standards bodies don't add device types just because they can. They add them when there's a clear demand signal from manufacturers and consumers. Right now, that signal is barely a whisper.
It's a chicken-and-egg problem. No one builds weight sensors because there's no standard, and there's no standard because no one builds weight sensors.
The way this usually breaks is that some company builds a compelling product using a proprietary protocol, it sells well, and the standards body says "okay, now we need a standard for this." But I don't see who builds the compelling consumer weight-sensing product. The use cases are too niche.
Which leaves the DIY path as not just the only option, but probably the best option even if commercial products existed. You get exactly the configuration you want, you understand how it works, and you can repair it when something breaks.
You get the satisfaction of building something that didn't exist before. There's a particular joy in wiring up four load cells, calibrating them, and watching real weight data stream into your Home Assistant dashboard. It feels like you've added a new sense to your house.
The house that can now feel how heavy things are.
It's a proprioceptive upgrade for your garage.
If you've built something like this — a DIY weight-sensing shelf, or any load-cell project that reports to a smart home — we want to hear about it. Send us your build notes, your calibration tricks, your mechanical mounting solutions. We're at show at my weird prompts dot com.
If you haven't built one but this episode convinced you to try, send us that too. Especially if you discover that one of your Euroboxes is way heavier than you thought.
Thanks to Hilbert Flumingtop for producing. This has been My Weird Prompts.
We'll be back soon.
