Retrofit vs New Build

Designing a smart building from scratch - greenfield - is the easy case in theory. You can specify an open BMS on BACnet, run structured cabling and risers, place meters and sensors where they matter, choose interoperable equipment, and hand over a clean BIM model that can seed a digital twin. The marginal cost of readiness during construction is small; the marginal cost of the same capability bolted on later is large. A building conceived as a data platform is dramatically cheaper to make intelligent than one that was not.

Brownfield retrofit is the harder, more common case. You inherit whatever is there: a proprietary BMS a decade old, controllers that only speak a vendor dialect, few meters, no structured data, and asset records in a filing cabinet. You cannot easily trench cable through an occupied, operating building, and you must work around tenants and around plant you are not allowed to switch off. The constraints are real - but so are the techniques, because this is exactly the situation the industry has spent years learning to solve.

The key retrofit insight is that you rarely rip out the existing BMS - you wrap and extend it. Where an old controller speaks a proprietary protocol, a gateway translates it to something open (BACnet/IP, Modbus TCP or MQTT) so its points become readable. Where cabling is impossible, wireless sensors - LoRaWAN, Zigbee or other low-power radios - add occupancy, temperature, humidity, CO2 and equipment vibration without pulling a single new wire, running for years on a battery. Where the utility meter is a black box, non-invasive sub-metering (clamp-on CT sensors, pulse counters) adds granular energy data on individual circuits and plant.

Above the hardware, an IoT platform ingests everything - old BMS points and new wireless sensors alike - into a time-series historian, and a data model such as Brick or Project Haystack tags each point with meaning so analytics can run across a messy, mixed estate. On top sit fault detection and diagnostics (FDD) and energy analytics that need no new plant at all - they simply make sense of data the building was already producing. This layered, additive approach is why a fifteen-year-old building can host genuinely modern analytics and, eventually, a lightweight twin, without a rebuild.

The mental shift that makes retrofit tractable is to stop thinking about the old BMS as an obstacle and start treating it as a data source you happen to already own. It is already sensing temperatures, running plant and logging alarms; the retrofit job is to liberate that data, add the points it is blind to, and give the whole lot meaning. Where you cannot reach a signal at all, a cheap wireless sensor fills the gap in an afternoon rather than a rewiring project. Where the twin needs geometry the building has no model for, a quick laser scan or even a tidy floor plan is enough to start - fidelity can grow later. Almost every retrofit constraint, looked at this way, has a pragmatic, additive answer that leaves the working building running while the intelligence is layered on top.

Retrofit works best as a staged journey, each phase funding the next (see the figure). Phase one is audit and quick wins: benchmark the energy use, meter the big loads, and switch on software-only analytics and FDD against the data you already have. These need little capital, often pay back inside a year, and build the owner's confidence - a stuck valve found, a schedule fixed, a demand peak shaved. Phase two integrates: wrap the old BMS with gateways, add wireless sensors where the blind spots are, and land everything on one platform with a shared data model. Phase three is analytics at depth: predictive maintenance, energy optimisation, occupancy-driven control. Phase four builds the twin: a live operational model, seeded from BIM if you have it or from a laser scan if you do not, connected to the now-rich data stream and serving named decisions.

The discipline is to never skip phase one. Owners who begin with the shiny twin and no baseline almost always overspend and underdeliver; owners who bank quick wins first arrive at the twin with a proven appetite, a funded team and a data foundation that actually supports it. Retrofit is a marathon run as a series of short, self-funding sprints.

Retrofit is mature, but it is not frictionless, and the failures follow a pattern worth knowing before you meet them. The commonest surprise is data quality: old sensors drift, meters get miscounted, and a point labelled one thing on the BMS turns out to measure another. Analytics run on bad data produce confident nonsense, so a retrofit almost always needs a phase of cleaning and validating what the building reports before trusting it - the unglamorous groundwork that separates a working system from a dashboard full of false alarms. Closely related is the integration surprise: a proprietary controller that will not expose its points, a licence needed to read the very data you already own, or a vendor lock-in that makes the existing system a walled garden. Discovering these early, through a proper survey, is far cheaper than discovering them mid-project.

Two more pitfalls deserve real respect. Cybersecurity gets worse, not better, when you bolt IoT onto an operational building: every new sensor, gateway and cloud connection is a new way in, and connected building systems have become genuine attack targets. Retrofit security - network segmentation, patching, credential hygiene - is specialist work and must be treated as such, with statutory and security sign-off left to qualified professionals, not assumed. Finally, the quiet killer is change management: a retrofit only delivers if the operators who run the building actually use the new analytics and act on the faults it raises. Technology dropped on an unwilling or untrained facilities team saves nothing. The best retrofits budget for the people and the process, not just the hardware - because the building that stays smart is the one whose operators were brought along.

Because phase one carries the whole momentum, it helps to know the moves that reliably pay. Sub-metering the big loads turns one opaque bill into an itemised map of where energy actually goes - usually the first time an operator sees it. Fault detection on existing BMS data surfaces the classic wasters: simultaneous heating and cooling, valves stuck open, dampers that never close, schedules running plant all weekend, sensors reading nonsense. Setpoint and schedule tuning - simply matching operation to real occupancy - is nearly free and often the single largest saving. Occupancy sensing on a floor or two reveals under-used space you may be heating, lighting and leasing for nothing.

None of these requires new plant or a rebuild; all of them make the existing building work as it should. They also do something subtler and vital: they generate the measured, before-and-after evidence that justifies the next phase. A retrofit that opens with three banked quick wins is a retrofit that gets its second budget - and that is how ordinary buildings, one pragmatic step at a time, become smart ones. The lesson to carry from all of this is that retrofit is less an act of construction than an act of patient, evidence-led sequencing, and it is a skill you can practise on almost any building you can walk into.