Energy analytics across 14 buildings and four BMS vendors

The situation

The portfolio team received one energy report per building per month, produced by four different service providers in four different shapes. Comparing buildings meant re-keying PDFs into a spreadsheet. An ESG commitment had made per-asset evidence a board-level requirement, and nobody could produce it.

The constraint

Three of the four BMS vendors treated data access as a paid integration, and one had no documented interface above the head-end. Facilities were — reasonably — unwilling to allow any system that could write a setpoint. And two buildings had no reliable IT network in plant rooms at all.

What we built

A survey that corrected the drawings

Six weeks of site work established what was actually installed, which controllers were reachable, and which points on the commissioning list had never been wired. Roughly one in nine documented points did not exist. Every later estimate depended on getting this right.

Read-only edge gateways

One gateway per building polls BACnet/IP and BACnet MS/TP over a router, plus Modbus TCP for the meters, and screen-scrapes a scheduled export from the closed system. Read-only by design and physically segmented from the control network. Local buffering means a WAN outage delays data rather than losing it.

Tagging, so a point means something

Each point is tagged against a Brick model: this sensor is the supply air temperature of that AHU, which serves those zones on that floor. This is unglamorous, it takes weeks, and it is the entire difference between a chart and an analysis.

Fault detection rules that name the asset

Simultaneous heating and cooling, valves stuck open, economisers not economising, equipment running outside occupancy, chillers short-cycling. Each rule produces an asset-level finding with an estimated annual cost, ranked. The facilities manager gets a work list, not a dashboard to interpret.

Baselines that survive finance

Weather-normalised regression baselines per building using an IPMVP Option C approach, so a savings claim can be defended when the finance team asks whether it was just a mild winter.

What changed

The first sixty days produced several hundred findings, of which roughly a fifth were worth acting on immediately — the largest being two air handling units that had run continuously since a controller replacement over a year earlier. Year-on-year energy use intensity fell by a double-digit percentage across the portfolio, most of it from operational corrections rather than capital work.

What we would do differently

We built the portfolio dashboard before the work-order integration. Findings sat in a tool the facilities team did not live in. Pushing findings into their existing ticketing system, which we did in month seven, should have been part of the pilot.

Outcomes

Portfolio energy use intensity

down double digits, year on year

Points under live monitoring

a few hundred → tens of thousands

Faults found in first 60 days

several hundred, roughly a fifth material

Client identity withheld under a mutual NDA. Figures are illustrative — rounded and directional, meant to show the shape of the change rather than an audited result. We will walk through the real numbers, and how they were measured, under NDA on a call.