Griotech Field Guide / Volume One
The Field Guide to Building Performance
The Performance
Chain Model.
a living operational model of connected building systems that traces causal performance relationships between them, attributing impact and cost at every node in the chain.
Foreword / Why not 'digital twin'?
The industry gave us
the wrong metaphor.
a twin is a copy, an identical replica. A building is not a manufactured product. What you need is not a mirror. What you need is a judge.
twin is a copy. An identical replica that mirrors the original. The word comes from manufacturing, where a digital twin is a precise virtual replica of a physical product. Same geometry, same materials, same tolerances. If a bearing wears in the physical product, the twin predicts it because it models the same physics at the same fidelity.
Buildings are not manufactured products. They are assembled, commissioned, occupied, modified, neglected, repaired, extended, and adapted over decades. The building you are operating today does not match the building that was designed fifteen years ago. Tenants have changed. Plant has been replaced. Controls have been rewritten, overridden, and rewritten again. The BMS has been patched by three different integrators. Half the sensors have drifted. A quarter of the valves don't close properly.
A digital twin of that building would need to mirror all of it: every override, every patch, every drifted sensor, every bodged control loop. And even if it could, what would it give you? A perfect replica of a messy building. A mirror held up to imperfection. The reflection would be accurate and completely useless.
What you need is not a mirror. What you need is a judge. Something that looks at the building, compares it to what it should be doing under current conditions, identifies the gaps, traces the consequences through connected systems, puts a cost on every one, and tells you what to fix first.
That is not a twin. That is a Performance Chain Model. And this guide explains what that means, why it matters, and how it works.
A mirror held up to imperfection is accurate and useless.
— Horace Cyrus, Founder/London · May 2026
Chapter One / The chain, not the box
What a building actually is.
a collection of connected systems whose performance is determined by the relationships between them, not by any single component in isolation. Four layers: Generation, Distribution, Air Handling, Zones.
he building performance industry has a habit of treating a building as a single entity. We talk about "building performance" as though a building performs. It doesn't. A building is a collection of connected systems that perform. And the connections matter more than the components.
Generation
The plant that produces heating and cooling. Boilers generating low-temperature hot water. Chillers generating chilled water. The energy enters the building here. This is the first link in the chain.
Distribution
The network that moves that energy around the building. Secondary CHW loops, secondary LTHW loops, pumps, valves, pipework. The energy travels here, and the efficiency of that travel determines how much of what was generated at source actually arrives at the point of use.
Air Handling
The systems that condition and deliver air to the occupied spaces. AHUs mixing fresh and return air, heating or cooling it, filtering it, delivering it at the right temperature and pressure. The energy is applied here.
Zones
The occupied spaces themselves. Fan coil units, VAV boxes, terminal reheat, radiators. The energy arrives here, and the occupant either feels comfortable or doesn't. This is where the chain's performance is ultimately judged.
The equipment is not independent. It is connected. And the connections carry cost.
Every building operates as a chain: Generation, Distribution, Air Handling, Zones. Energy is produced, moved, applied, and delivered. What matters is not any single link in the chain. What matters is how the links affect each other.
When a chiller's COP degrades, the symptom does not stay at the chiller. The degraded cooling propagates downstream. The secondary loop works harder to compensate. The AHU receives warmer chilled water, so supply air temperatures rise. The FCUs in the zones can't cool effectively, so occupants feel warm.
One fault. Four layers affected. Six systems compensating. The cost is not at the chiller. The cost is distributed across the entire chain. Any model that treats each piece of equipment as an independent entity is missing the point. The equipment is connected. The connections carry cost.
Chapter Two / Five scores, three layers
The language of performance.
five normalised scores arranged in three layers, computed at every node, continuously. The vital signs of a building. NUF and AER at the foundation, MEI and CEI as composites, OEI as the headline.
efore we can explain what a Performance Chain Model does, we need to establish the language it speaks. The chain communicates through five normalised scores arranged in three layers, computed at every node, continuously. Think of these as the vital signs of a building.
Layer one · The two foundational measurements
The foundation asks two questions of every system, continuously: how much energy is it consuming, and how efficiently is it converting that energy into useful work?
Normalised Usage Factor
How much energy is this system consuming relative to what it should consume under current conditions? The key phrase is "under current conditions." Normalised usage doesn't compare a system to a fixed benchmark. It compares the system to its own weather-adjusted regression baseline, built from heating and cooling degree days.
A score of 0.85 means the system is consuming 85% of its expected energy — better than baseline. A score of 1.3 means 30% more than expected. A cold snap doesn't inflate the score because the baseline shifts with the degree days. A mild day with elevated consumption produces a genuinely high reading, not seasonal noise.
Adjusted Efficiency Rating
How efficiently is this system converting energy input into useful output, adjusted for operating context? Where normalised usage tells you how much a system is consuming, adjusted efficiency tells you how well it's converting that consumption into useful work.
A chiller might have normal usage (consuming the right amount) but degraded efficiency (producing less cooling per unit consumed). High usage and low efficiency means: consuming too much and not converting it well. Normal usage and low efficiency means: consuming the right amount but getting less output — likely a partial fault or degradation.
Layer two · The two composite dimensions
The second layer takes the foundational measurements and produces two composite scores: one for mechanical effort, one for comfort delivery.
Mechanical Effort Index
Mechanical effort is a composite of NUF and AER. It answers a single question: how hard is this system working relative to how well it's converting that work? A system with high usage and high efficiency is working hard but productively. A system with high usage and low efficiency is working hard and wasting energy in the process. That is a fault.
Comfort Effectiveness Index
Comfort delivery is computed independently from the energy metrics. It draws on temperature tracking accuracy, pressure stability, and setpoint adherence across the served zones. It captures the outcome the occupant actually experiences.
This is the metric that answers the question the building exists to serve: are the people inside comfortable? Everything else — every watt consumed, every valve opened, every fan run — exists in service of this number.
Three layers. NUF and AER at the foundation. MEI and CEI as composites. OEI as the headline. Computed at every node. Continuously.
Overall Effectiveness Index
Overall effectiveness is composed from MEI and CEI. It brings the energy dimension and the comfort dimension together into a single number representing the system's total performance position. This is the score that appears at the top of every node in the chain. The number that turns green, amber, or red.
The composition uses adaptive weighting based on data availability and confidence. The confidence score is always visible, so the reader knows how much to trust the number. An OEI of 39% can be decomposed. Every headline is traceable.
Chapter Three / And none of them are this
Four meanings of
digital twin.
the four technical implementations vendors call 'digital twin': a physics simulation, a data ontology, a statistical baseline, a 3D render. Four problems. Four solutions. One marketing term.
efore we explain what a Performance Chain Model is, we need to be honest about what "digital twin" has become. The term is used by at least four categories of vendor in the building performance space, and each means something different.
Type 1 · The physics simulation
Some platforms build physics-based expert systems using engineering equations and logical statements to simulate how each piece of equipment should operate, comparing that simulation against actual BMS data to detect faults.
This is the most intellectually rigorous use of the term. It is also equipment-centric. Each piece of equipment gets its own model. What is absent is the relationship between them. The twin knows that the chiller is underperforming. It does not trace the cost of that underperformance through the distribution network, into the AHUs, and down to the zones.
Type 2 · The data ontology
Other platforms use "digital twin" to describe a data integration layer — a normalised schema that unifies data from diverse building systems into a single model, with analytics and fault detection sitting on top.
This is useful infrastructure. But it is not a model of how the building behaves. It is a model of how the building's data is structured. The twin doesn't reason about the building. It organises information so that other tools can.
Type 3 · The statistical baseline
Some platforms build machine-learning consumption baselines. The system learns what energy consumption looks like under normal conditions and flags when actual consumption deviates. This is closer to what most people imagine when they hear "digital twin" — it predicts and compares.
But it operates at the whole-building or whole-system level. It knows that consumption is elevated. It does not know why, which system is responsible, or what the upstream and downstream costs are.
Type 4 · The 3D render
Other platforms create geometric, visual replicas of buildings — often derived from BIM models — with sensor data pinned to the 3D geometry. Compelling to look at. Also the least useful for operational decision-making.
A 3D model shows you where equipment is. It does not show you how equipment affects each other. Spatial co-location is not causation.
Four technical implementations. One marketing term. Zero consensus on what it actually means.
This is the environment in which the Performance Chain Model was developed. Not as a better digital twin, but as a deliberate departure from the concept entirely. A term that differentiates nothing distinguishes no one.
Chapter Four / Operational · Causal · Living
The Performance Chain Model.
the model at the centre of a continuous performance audit. Not a digital twin. It does not replicate the building. It judges the building.
he Performance Chain Model is the term we use for the model that sits at the centre of a continuous performance audit. It is not a digital twin. It does not replicate the building. It judges the building.
The PCM has three properties that distinguish it from every implementation of digital twin described in the previous chapter.
It is operational
The PCM tracks the live performance of real systems in real time. Not design performance. Not commissioning baselines. Not nameplate ratings. Actual, observed, current performance, normalised for weather, occupancy, and operational context.
Every system in the chain carries the five live scores: NUF and AER at the foundation, combining into MEI, alongside independently computed CEI, with both rolling up into OEI. These are not static values configured at setup. They are computed continuously from live BMS data, normalised against weather-adjusted regression baselines that evolve as the building changes.
It is causal
This is what no digital twin implementation does. The PCM doesn't just score each system independently. It traces causal relationships between them. When the chiller's AER drops, the PCM asks: what is happening to the NUF on the secondary CHW loop? Is it compensating? What about the AHU — is its CEI degrading because the chilled water is warmer?
The PCM traces the impact upstream (what is causing this fault?) and downstream (what is this fault causing?), reading the scores at every node along the chain. And at every node, the cost is attributed using plant ratings.
Upstream: £17.82. Local: £4.17. Downstream: £3.85. Total: £25.84. One fault, four layers, six systems, one cost figure.
It is living
The PCM is not configured once and left. It adapts. HDD/CDD regression baselines evolve as the building's weather-energy relationship changes. If a new boiler is installed, the baseline adjusts over time as the regression accumulates data under the new conditions.
A digital twin configured from design data is frozen in time. It models the building as it was intended to be. The PCM models the building as it is.
Chapter Five / A fault walks downstream
How the chain traces impact.
to traverse the chain in two directions when a fault is detected — upstream to find the root cause, downstream to attribute the propagated cost — at every node, in pounds.
he "chain" in Performance Chain Model is literal. It refers to the directed graph the PCM builds from the building's system topology: Generation, Distribution, Air Handling, Zones. When a fault is detected at any node in the chain, the PCM performs two traversals.
The upstream traversal
The upstream traversal asks: is this fault caused by something further up the chain? If an AHU's supply air temperature is elevated, the PCM checks whether the upstream CHW distribution is delivering at the expected temperature. If it is, the fault is local to the AHU. If it isn't, the PCM follows the chain further upstream to the generation plant.
This is the distinction between a localised fault and a propagated fault. Traditional fault detection sees both as AHU faults. The PCM distinguishes them, because the recommended action is completely different: fix the AHU damper (local) versus fix the chiller (upstream, with the AHU as a symptom).
The downstream traversal
The downstream traversal asks: what is this fault causing in connected systems? If the chiller's efficiency is degraded, the PCM traces the effect forward: CHW Secondary loop compensating (£1.80/day), LTHW Secondary affected (£1.80/day), CHW Primary demand elevated (£3.20/day). Further downstream: AHU supply temperatures drifting (£4.17/day), FCUs overdriving (£3.85/day). Comfort degrading across four zones.
At every node, the PCM records three things: the cost in pounds, the comfort impact in percentage points, and the excess energy in kWh. The total of all nodes gives you the full impact of a single fault across the entire chain.
Traditional FDD: "Chiller low COP. Priority: High." The PCM: "Chiller low COP. Upstream £17.82. Local £4.17. Downstream £3.85. Total £25.84. Comfort −3.2%. Excess 148 kWh. Scope: 6 systems."
The first is a detection. The second is evidence. Detection is a label. Evidence is a chain with a price tag.
Chapter Six / A worked example
What this means in practice.
to read the chain end-to-end on a real building: ten systems, twenty-six faults, every node carrying an OEI score and a cost figure. The chain in action.
commercial building. Ten systems monitored. Twenty-six active faults. The Performance Chain Model has been running continuously, scoring every node, tracing every connection. The Impact Chain, the visual representation of the PCM, shows four columns: Generation, Distribution, Air Handling, Zones. Each system is a node with a live OEI score, fault count, and cost figure.
Reading the chain
Generation: CHW Primary at 84%, £187 in recoverable value, one fault. LTHW Primary at 92%, performing well.
Distribution: CHW Secondary at 58%, £49, two faults. LTHW Secondary at 61%.
Air Handling: AHU-01 at 39%, £73, seven faults. The worst-performing system in the chain. AHU-02 at 64%, £25, four faults.
Zones: Floor 3 FCUs at 48%, £33. Floor 4 FCUs at 53%, £21. Floor 2 FCUs at 47%, £18. Floor 1 FCUs at 66%, £0, performing adequately.
The chain reads left to right: generation produces, distribution moves, air handling applies, zones receive. Where the colour shifts, that is where performance degrades.
Tracing the root cause
AHU-01 is the worst node. But is it the root cause? The PCM's upstream traversal checks. CHW Primary is at 84% with a chiller low COP fault. That degradation is propagating downstream through CHW Secondary (58%) into AHU-01 (39%). The AHU also has its own local faults: fresh air damper stuck open, unstable pressure control. It is both a victim of upstream degradation and a source of local faults.
The downstream traversal from AHU-01 shows the effect on its served zones: Floor 3 FCUs at 48%, Floor 2 FCUs at 47%. Both below 50%. The elevated supply temperatures from the AHU are forcing the FCUs to compensate. Total downstream cost from the chiller fault alone: £218 distributed across every downstream node.
The Impact Chain is not a system diagram. It is a financial map. Every node carries a price tag. The chain shows where the money is going. This is what the PCM makes visible: not a single fault on a single piece of equipment, but the full consequence of interconnected degradation across the entire chain.
Closing / Relationships, not replicas
Relationships, not replicas.
to abandon the idea that the building must be replicated, and replace it with the idea that the relationships between its connected systems must be understood, judged, and costed.
Buildings are not products to be replicated. They are networks of connected systems to be understood, judged, and costed.
The building performance industry has spent a decade reaching for the digital twin as its central metaphor. The appeal is understandable. A twin implies completeness, fidelity, a virtual replica so accurate that it can predict and prevent. It is a comforting idea.
It is also the wrong one. Buildings are not products to be replicated. They are networks of connected systems to be understood. The value is not in mirroring the building's state. It is in judging the building's performance, tracing the consequences of degradation through the system chain, and putting a cost on every link.
That is what the Performance Chain Model does. It is operational, tracking live performance normalised for weather and context. It is causal, tracing impact upstream and downstream through the Generation, Distribution, Air Handling, Zones chain, with cost attributed at every node. And it is living, learning from observed behaviour as the building changes, not frozen in the assumptions of a design model.
When a chiller's efficiency drops and the downstream AHU's usage rises and the zone's comfort delivery degrades, the chain is telling you something. Not as a correlation. As a cause-and-effect trace with a pound sign at every step. A digital twin would mirror that same building and show you the same temperatures. The PCM tells you what those temperatures are costing, why they're wrong, and what to do about it.
We did not build the Performance Chain Model because we thought digital twins were bad. We built it because we thought the question was wrong. The question was never "can we replicate the building?" The question was "can we understand the relationships between connected systems well enough to judge performance, trace impact, and recover value?" The Performance Chain Model is our answer.
See the chain in action.
In thirty minutes.
A walkthrough of the platform. The Performance Chain Model rendered live for a real building. Costed faults, traced through the system chain, briefed in plain language.
Prefer to read it on paper?
View supporting research and methods
Explore the ABIE engineering engines, supporting evidence and methodology and building applications.
Field Guide Series