Good Building Intelligence Begins with Good Data
Why KyoSensa chose Tector as its sensing foundation, and why dependable hardware is only the beginning of effective building monitoring.
Every building monitoring system ultimately depends on something very simple: the quality of the information entering it.
Machine learning, sophisticated analysis and predictive models cannot compensate for unreliable measurements. If sensor data is inconsistent, intermittent or poorly suited to the environment in which it is being collected, every layer built above it becomes less trustworthy. For KyoSensa, choosing the right sensing technology was therefore one of the most important technical decisions we could make.
We chose Tector.
The decision was not based on finding a sensor with the longest feature list or the most impressive specification sheet. We wanted a sensing platform that had already demonstrated that it could operate in real buildings, under real environmental conditions, for long periods of time. Just as importantly, we needed a technology that could form a stable foundation for something considerably broader.
KyoSensa is designed to understand buildings over time.
That requires reliable measurements first. What happens to those measurements afterwards is where KyoSensa begins.
Sensors are not interchangeable
It is easy to think of moisture sensors as relatively simple devices. Install them, collect readings and wait for something to cross a threshold.
In practice, long-term building monitoring is much more demanding.
Sensors may need to remain inside roofs, walls, structural timber or other difficult-to-access locations for years. They have to survive changing temperatures and humidity, communicate reliably through the building, and continue generating sufficiently consistent data for long-term comparisons to remain meaningful.
This becomes even more important when the objective moves beyond simple alarm thresholds.
A system such as KyoSensa needs to determine how a building normally behaves before it can reliably identify behaviour that is unusual. That means comparing months and eventually years of measurements. Small inconsistencies that might be insignificant when looking at a single reading can become far more important when those readings form part of a machine-learning model.
The quality of the sensor network therefore affects the quality of everything that follows.
Why we chose Tector
Tector gave us several characteristics we considered essential.
Its wireless sensors are purpose-built for long-term moisture monitoring in timber structures, flat roofs and other building assemblies. The current generation uses LoRaWAN connectivity and measures parameters including moisture content, relative humidity and temperature. Tector states a service life of up to 25 years under its specified operating conditions, while its current generation carries certifications for markets including Japan and Australia.
For KyoSensa, however, one factor carried particular weight: the technology had already moved well beyond laboratory demonstrations.
Tector sensors are being used on real construction projects, occupied buildings and culturally significant structures where monitoring has practical consequences.
During the renovation of Copenhagen City Hall Tower, for example, Rambøll used continuous Tector measurements to follow moisture conditions and adjust heating and dehumidification as the structure dried.
In the Faroe Islands' first cross-laminated timber building, continuous measurements were used during construction for quality assurance and to inform decisions about when timber components were sufficiently dry for subsequent work. Sensors also remained in parts of the completed building for long-term monitoring.
Tector technology has also been deployed in the restoration of Notre-Dame in Paris, where sensors allow moisture conditions within difficult-to-access oak structural elements to be followed over time.
A very different example can be found in Copenhagen, where sensors were installed across a series of multifunctional school roofs incorporating gardens, solar panels, outdoor facilities and other penetrations. During construction, the system detected rainwater entering through an exposed assembly, allowing the project team to identify the affected area and determine whether further drying was required before the roof was closed. The monitoring infrastructure is intended to remain in place during operation, providing continued visibility into a roof where finding future water ingress after the fact could otherwise be difficult and disruptive.
These projects are very different from one another. That is precisely why they matter.
They demonstrate that the underlying sensing technology can function across construction, long-term building operation, modern mass timber, complex flat roofs and heritage conservation. For KyoSensa, that provided the confidence we needed to build a much more sophisticated intelligence layer around it.
Reliable measurement is the beginning, not the outcome
A sensor can tell us what it measured.
A building owner or facilities manager usually needs to know something more important:
What does that measurement mean for this building?
A moisture reading viewed in isolation has limited context. The same value may be completely normal in one location and unusual in another. A rapid increase might indicate water ingress, or it might correspond perfectly with heavy rainfall and the normal response of that part of the structure. Elevated moisture that rapidly returns to normal can have a very different significance from a smaller change that persists for weeks.
The relationships between sensors can also matter.
If every sensor in one part of a building responds similarly to a weather event except one, that difference can be significant. If two nearby sensors historically behave together and then begin to diverge, the relationship itself becomes useful information. If an area that normally dries within several days suddenly takes considerably longer, the recovery pattern may deserve attention even if no conventional alarm threshold has been crossed.
This is the problem KyoSensa is being built to solve.
From sensor network to Building Fingerprint
KyoSensa continuously develops what we call a Building Fingerprint.
Rather than treating every measurement as an isolated event, the platform learns how the building behaves as a system.
It considers historical sensor behaviour, seasonal variation, rainfall, environmental conditions, drying and recovery patterns, persistence of elevated readings and the relationships between different sensors and areas of the building. In Japan, the platform can also incorporate local seismic information so that unusual behaviour following significant ground movement can be assessed with additional context.
Over time, this creates a model of expected behaviour for the individual property.
KyoSensa can then compare what is actually happening with what it has learned would normally be expected.
That distinction is important.
Traditional monitoring is largely based on asking whether a measurement has exceeded a predefined value. KyoSensa can also ask whether the building is behaving differently from itself.
Those are not the same question.
A value can remain inside an acceptable range while its rate of change, persistence or relationship with neighbouring sensors becomes unusual. Conversely, an elevated reading may be entirely consistent with a known environmental event and normal historical behaviour.
The objective is not to produce more alerts. It is to produce better information.
Building intelligence needs context
KyoSensa therefore combines sensor measurements with information beyond the sensor itself.
Weather is an obvious example. Rainfall intensity and timing can dramatically alter the significance of a moisture change. The system can examine how a building reacted to similar rainfall previously, how quickly individual locations responded and how long they normally required to recover.
The same principle applies across the building.
KyoSensa analyses individual sensors, zones and relationships between them. It examines expected versus observed behaviour, changing recovery curves, unusual persistence and emerging deviations from the building's established patterns.
Our Insight Engine brings those analyses together and determines which changes deserve further attention.
Importantly, KyoSensa is not intended to remove people from the process. Building management involves consequences, costs and physical conditions that software cannot always see. Significant findings can therefore be surfaced for human review, giving building owners and managers both automated analysis and professional oversight rather than forcing them to interpret streams of raw sensor data themselves.
The technology does the continuous observation. People remain responsible for the important decisions.
A different level of monitoring
This becomes particularly valuable when monitoring moves from a handful of sensors to an entire portfolio.
A property manager responsible for multiple buildings should not need to spend each morning opening sensor graphs and deciding whether anything looks unusual. Nor should an owner of an important heritage structure need to become an expert in moisture dynamics to benefit from long-term monitoring.
KyoSensa is designed to provide that interpretation.
The platform can bring together buildings, sensor locations, environmental information, alerts, analysis history and reports within a single operational environment. It can distinguish between routine changes and behaviour requiring attention, maintain a history of what occurred and provide the reasoning behind significant insights.
For service providers and building-management organisations, the same architecture can extend across large numbers of properties while maintaining a Building Fingerprint for each individual structure.
A centuries-old machiya in Kyoto should not be expected to behave like a modern commercial roof. A warehouse should not be assessed as though it were a timber temple. Even two apparently similar buildings can respond differently because of orientation, materials, drainage, ventilation, repairs, occupancy and years of previous weathering.
KyoSensa learns those differences rather than assuming they do not exist.
Why this matters for prediction
Long-term monitoring becomes increasingly valuable as the historical record grows.
The first objective is to detect meaningful change earlier.
The next is to understand why it is occurring.
Ultimately, the objective is to recognise developing behaviour early enough to estimate where conditions may be heading before a conventional alarm would have been triggered.
That is where the combination of dependable sensing and building-specific machine learning becomes particularly powerful.
A predictive system cannot simply be told that moisture above a particular number is dangerous and then claim to understand a building. It needs to learn normal behaviour, compare similar events, identify changing relationships, understand how quickly conditions normally recover and recognise when the current trajectory no longer resembles the past.
That requires data collected consistently over time.
It is one of the reasons our choice of sensor partner mattered so much.
Choosing a monitoring partner, not simply a sensor
For a building owner, contractor or facility manager considering long-term monitoring, the hardware is important. It should be reliable, appropriate for the structure and proven under real operating conditions.
But hardware should not be the end of the evaluation.
The more important question is what the monitoring system will enable once those sensors are installed.
Who will interpret the information?
Will the system understand the building's history?
Can it distinguish a normal environmental response from an emerging anomaly?
Can it compare sensors and zones rather than assessing each device independently?
Can weather, seismic events and other external information be incorporated into the analysis?
Will the system become more useful after several years of operation, or will it simply continue presenting the same graphs?
These questions shaped KyoSensa from the beginning.
We did not want to manufacture another moisture sensor simply so that every component carried our name. We wanted to use strong sensing technology as the foundation for a larger system whose value increases as it learns.
Tector provides that foundation.
KyoSensa turns the resulting measurements into a continuously developing understanding of the building itself.
That distinction is central to what we are building.
Reliable sensors tell us what is happening at a particular point in a structure. KyoSensa connects those observations across time, weather, events, locations and the behaviour of the rest of the building.
Sensors measure. KyoSensa learns.