KNX was born in the nineties as Europe's open protocol for building automation. The promise was simple: any certified KNX device should talk to any other certified KNX device, regardless of manufacturer. In theory it makes sense. In practice, thirty years later, the result is very different.
Today I'm going to tell you three real cases I've lived through in recent years. The main one is a shoe factory in Elche that spent two years on the verge of shutting down because of a badly designed KNX installation, until we decided to literally rip everything out and start from scratch with Shelly and a Raspberry Pi. But first, let me explain why these situations are far more common than they look.
The myth of the open standard
KNX is certified by the KNX Association, based in Brussels. The name sounds like a neutral consortium. The reality is that to work as a professional integrator you need ETS, the official programming software, which costs around €4,000 for the professional version and is 100% proprietary. Every KNX project lives inside an ETS file — if you lose the file, or if the original integrator doesn't hand it over at the end of the installation, you can't reprogram anything without starting from scratch.
No matter how "open" the protocol is on the wire, the entry door to the ecosystem is expensive proprietary software. And that already starts to contradict the word "standard".
The second trap is manufacturer fragmentation. Gira, Jung, ABB, MDT, Siemens, Schneider, Theben, Hager... they all sell "KNX certified" products. But each one implements its own extensions, its own application programs, and has its own catalogue of references. When you mix manufacturers in the same installation, incompatibilities show up in the corners of the project: a relay output that won't obey a binary input, a thermostat that ignores bus events when the setpoint changes, a DALI gateway that responds out of sync. Technically everything meets the standard. Functionally, nothing works together.
The third trap is stock. KNX isn't consumer hardware — it's professional integration hardware. When a reference goes end-of-life (and it happens constantly), you can spend weeks looking for a specific part across Europe. And when you find it, it costs twice what it did when you first did the install.
When one of these three problems catches you off-guard, you don't have a home automation installation — you have an expensive time bomb.
The Elche case: two years of chaos
In 2019 a shoe factory in Elche contacted us about a problem they had been trying to fix with their original installer for two years. They had invested close to €60,000 in a "turnkey" KNX system that controlled:
- Interior lighting of the production floor
- Exterior and parking lot lighting
- Control of the sole heat-setting ovens
- Office HVAC
All integrated, all automated, all "compliant with the European standard". The problem is that from day one, nothing worked right. Lights came on by themselves at night. The ovens — which are the most critical piece of production, because without them there are no shoes — shut down for no apparent reason and, worst of all, there was no way to manually restart them without calling the integrator. The factory stopped for 4 to 8 hours every time it happened.
The client would ask for one thing, the installer would do another, and whenever something failed the answer was always the same: "it's the French manufacturer's fault, the part has a bug, we're waiting for new firmware". Two years of new firmware that never arrived. (I say "French manufacturer" to keep brand names out of this — anyone who works in this sector will know who I mean.)
They called us because they were on the verge of collapse. Every outage stopped the production line. Soles that went into the oven while it was shutting down came out half heat-set and had to be thrown away. Operators spent half their time on the phone trying to reach the original integrator, who had stopped picking up.
When a factory loses 4 to 8 hours of production every time the lighting system has a hiccup, what you have isn't a home automation installation. It's an expensive time bomb.
Technical diagnosis
The first thing I asked for was the original ETS project file. It took them a week to get it to me — the integrator had lost it "in a computer upgrade". When we finally opened it, we found what I already suspected: a KNX architecture with close to eighty devices from four different manufacturers, with bus areas that didn't follow the recommended topology, feedback loops between actuators and sensors, and an oven management system implemented as a cascade of communication objects that depended on real-time bus state.
In plain English: the system worked badly because it was badly designed. The ovens had their safety logic — maximum temperature, overheating detection, fault shutdown — delegated to the KNX bus. Any bus collision or delay translated into a shutdown command that the oven obeyed without knowing why. And because the integrator had programmed everything with interlocking group objects, restarting an oven literally required reprogramming from ETS.
This architecture should never have left the drawing board. But when KNX lets you touch everything from the bus, there's a dangerous incentive to put critical logic on the bus. No serious industrial technician would put the safety interlock of an oven on a bus shared with the parking lot lights. And yet, there it was.
The solution: Shelly Pro + Raspberry Pi
The decision was radical: rip everything out. Literally everything. Devices, actuators, sensors, gateways. We recovered what could be resold on the secondary market and the rest went to scrap.
In its place we installed a much simpler and much more robust architecture:
- Shelly Pro 4PM for every lighting group. They're modular DIN actuators, four channels each, per-channel power metering, and WiFi + Ethernet communication. They work without needing a central server: if the coordinator dies, each Shelly keeps its local logic running.
- Shelly Pro EM-50 for three-phase power metering in every critical zone, with real-time reporting to an in-house database.
- The ovens got their own dedicated panel back, with hardwired safety interlocks and a physical reset relay. The only integration with the general system is a status signal (on/off) and an emergency-stop command, both over dry contacts. No shared buses for critical logic.
- A Raspberry Pi 4 running Home Assistant as the coordinator for the whole system. Automation scripts in YAML, automatic backups to a client-owned NAS, and a web interface the floor manager accesses from his phone.
- A 4G modem for fallback for when the factory WiFi had problems, which happened often because of electromagnetic noise from the machinery.
The total cost of the full migration project was €6,200 in hardware plus 8 days of work from two technicians. Less than one-tenth of the cost of the original KNX system.
That installation has been running since 2019. Zero outages attributable to the system. The ovens haven't shut down on their own even once. When a Shelly has failed (it's happened twice in six years), the client bought a new one on Amazon, swapped it in twenty minutes and reconfigured it by cloning the config via Home Assistant. No external technician needed.
Other cases I've seen
Elche isn't an isolated case. In recent years I've seen variants of the same problem across very different sectors. These are three more I remember well:
Pharmacy in Madrid — discontinued part, refrigerators at risk
€80,000 invested in a Siemens KNX system controlling lighting, HVAC and the back-room refrigerators where temperature-sensitive medication is kept. In year three, the specific thermostat reference went end-of-life and there wasn't a compatible one in stock anywhere in Spain. Two weeks with refrigerators in manual mode — checking temperatures every hour with a pocket thermometer for fear of losing a batch of insulin — while we hunted for a replacement part on German eBay at three times the original price.
The final fix was to replace everything critical (refrigerators + prep-room HVAC) with conventional industrial controllers and a Shelly H&T as a backup monitor that alerts the owner via Telegram if any temperature drifts out of range. The client kept KNX just for the lighting, because it was already installed and working — the rule holds there: if it works, don't touch it.
Apartment building in Alicante — phantom alarms
KNX system controlling garage access, common lighting and perimeter alarms. Every night, at random, one of the alarms would fire by itself. The residents called the integrator, who charged for the visit, blamed "a lightning strike", and left. Two weeks later, another phantom alarm. And another. And another.
The original integrator claimed it was "electromagnetic noise on the bus" and proposed a €1,200 filtering gateway plus a visit fee to install it. When they called us for a second opinion, the real diagnosis took twenty minutes: the system's IP gateway had a known bug the manufacturer had fixed in a firmware update — but the new firmware required updating ETS to a version the integrator didn't own and refused to buy. Fix: we replaced the gateway with a Shelly Plus i4 that sends events to Home Assistant, and the false positives were gone the same day. Total cost: €45 in hardware and 2 hours of work.
Tech consultancy offices — impossible integration
A modern company with a KNX system that they wanted to integrate with Google Calendar, so meeting rooms would automatically cut their lighting and HVAC when they'd been empty for fifteen minutes and had no upcoming bookings. On paper, a trivial integration. In the real KNX world, impossible without expensive gateways, ETS programming and every logic change billed separately.
We migrated the meeting rooms to Shelly + Home Assistant + HACS, and the company's in-house IT team programmed the Google Workspace integration in one afternoon, without calling any outside consultant. The original KNX integrator had been asking them for money for a "feasibility study" for six months — a study they never delivered.
When KNX does make sense
I don't want this article to sound like a blanket attack. KNX still has scenarios where it's the right choice:
- Large luxury villas with a single trusted integrator who'll maintain the system for years and where budget isn't the problem. Here the complexity of the bus, the expensive software and the specific parts are justifiable because the client wants an elegant system, Gira or Jung touch panels in every room, and doesn't mind paying for it.
- Projects where KNX is already installed and working well. If you inherit a working installation, don't touch it. The cost of migrating is almost always higher than the benefit. Layer on top if you need more functionality, but don't rebuild what works.
- Specific integrations with certain European HVAC manufacturers that only offer native KNX interfaces, where gateways to other protocols are expensive or unstable.
- Certified buildings where a KNX system's formal documentation makes energy-efficiency audits or LEED/BREEAM certifications easier.
In every other case — and especially in industrial environments where downtime is expensive and simplicity of replacement is critical — there are better alternatives.
What to use instead of KNX
My technical recommendation by project type, no fluff and with no loyalty to any manufacturer:
- Regular residential home (< 200 m²): Standard Shelly + Home Assistant. Per-point cost: €25–40. Web-based configuration. Components available on Amazon with 24h delivery.
- Luxury home or large villa: Shelly Pro on DIN rail + Home Assistant on an Intel NUC with SSD. Same philosophy as residential but with professional DIN actuators and a physical server backup. Per-point cost: €40–80.
- Offices and commercial spaces: Shelly Pro + ESPHome (for custom sensors) + Home Assistant + integration with corporate APIs. The added value here is the ability to integrate with Slack, Google Calendar, Microsoft Teams, etc.
- Industry: Conventional PLC controllers (Siemens S7, Schneider M2xx) for critical safety logic. Shelly Pro + Home Assistant for the supervision layer and general lighting. Never mix the two on the same bus. The rule is simple: safety goes through physical cable, supervision goes over IP.
- Agricultural / greenhouses / off-grid: Same pattern as industry, but with special attention to power consumption because these systems usually run on off-grid solar. Here I've successfully used Raspberry Pi + Node-RED + Shelly Plus for irrigation, ventilation, temperature and water-level control. That's what I did on the Murcia mountain project I describe on my projects page.
Conclusion
KNX isn't dead. It has legitimate use cases. But it stopped being the default answer at least a decade ago, and traditional European manufacturers still haven't caught on.
When a client asks me for an industrial automation system, my first question isn't "KNX or not KNX?" but "what's your plan for the next ten years?". If the answer includes a single outside integrator, budget for expensive spare parts and time to wait for firmware, KNX might make sense. In any other case, you're probably better off with a combination of Shelly, Home Assistant, ESPHome, and a bit of industrial common sense.
The Elche client has been running six years without problems. He hasn't had to call anyone. His factory hasn't stopped. And when a part fails, he buys a new one on Amazon for under €100 and swaps it himself in twenty minutes. That's the kind of home automation I want to build.