What the studies really say about BMS optimisation savings

Every integrator who sells BMS service is eventually asked the same question: how much will this save? It is tempting to reach for a headline percentage from a brochure. A better answer starts with what the large published studies actually measured, how many buildings they covered, where those buildings were and how wide the spread of results was.

This article walks through the main evidence, the figures that are out of date, and a way of talking about savings that a technical client - or a sceptical finance director - will trust. None of it is a promise for any one building. The honest message is that most existing systems carry recoverable waste, the size of it varies widely, and the only way to know is to measure before and after.

The largest dataset - and why 16% is out of date

The best-known figure in the industry is a median saving of 16% from commissioning existing buildings, with a payback of about a year. It comes from a 2011 study by Lawrence Berkeley National Laboratory (LBNL) in the US, and it still turns up in sales material.

LBNL re-ran that work in 2020 with a much bigger dataset - nearly 1,500 North American buildings (Crowe et al. 2020). The combined median whole-building saving from existing-building commissioning came out at 6.4%, with an interquartile range of 3.4-12.4% across 446 projects, and a median simple payback of 1.7 years. On the newer 2018 data alone the median was 6.0%; on the older 2009 data, re-analysed, it was 10.0%. In both samples some projects saved over 50% and some saved almost nothing.

So the up-to-date headline is "about 6%, typically somewhere between 3% and 12%", not 16%. That is still a worthwhile return for work that is mostly schedules, setpoints and sequences - but it is a different conversation.

What the large studies measured - and the modelled ceiling - Whole-building energy savings, % (medians)
Recommissioning existing buildings - 446 projects (measured)6.4% (IQR 3.4-12.4%)
Fault detection and diagnostics (FDD) - after two years (measured)9%
Potential found in BMS trend data - 151 buildings (potential)12%
Optimised controls across the whole stock (modelled)29%

Medians, not promises - individual results ran from close to zero to over 50%.

The measured medians sit well below the modelled ceiling - that gap is the untapped potential. All North American. Source: Crowe et al. 2020 (LBNL); Smart Energy Analytics Campaign, cited in Crowe et al. 2023 (LBNL); Katipamula et al. 2021 (PNNL); Fernandez et al. 2017 (PNNL, DOE summary). Flags: North American, Top bar modelled.

Monitoring, trend data and the modelled ceiling

Three other figures belong alongside the recommissioning median:

Read together, the gap between the 6-9% actually measured and the 29% modelled is a fair way to describe untapped potential - without claiming any of it for a particular site.

What actually gets fixed

The LBNL dataset also records what the commissioning teams did. Across 3,695 measures from 503 projects, five measure types made up 95% of the total: scheduling, other operation and control changes, advanced resets, setpoint changes and sequence-of-operation changes. 87% of measures were HVAC-related. That list is, almost word for word, the work a controls integrator does on a service visit.

The study also shows that building type matters, though the samples get small quickly. Median savings ran from 3% in hotels to 16% in public order and safety buildings, with laboratories at 14% and data centres at 11% - the latter from only four projects.

Recommissioning savings by building type - Median whole-building energy savings, %
Lodging (hotels)3%
All building types combined (n = 446)6.4%
Data centres (n = 4)11%
Laboratories (n = 28)14%
Public order and safety16%

The lowest and highest segments and three in between - small samples (n = 4) say little on their own.

Building type changes the median, but small samples (data centres, n = 4) say little on their own. North American. Source: Crowe et al. 2020 (LBNL), existing-building commissioning medians by market segment. Flags: North American, Measured.

Where the faults are

A separate LBNL study looked at what FDD tools actually report in day-to-day operation: 317 buildings, 3,660 air handling units and 53,865 terminal units (Crowe et al. 2023). On any given day, 40% of AHUs and 30% of terminal units had at least one fault. The most common AHU fault - supply air temperature not tracking its setpoint - affected 55% of units.

The faults that were both common and persistent cluster in a handful of places: supply air temperature and static pressure setpoints, the economiser (free-cooling) sequence, the supply air sensor, coil valves and airflow sensors. That gives an engineer a short list of where to look first. The study covers AHUs and terminals only - there is no dataset of similar scale for chillers, boilers or heat pumps, and none for the UK.

How common faults are in air handling and terminal units - Share of units, %
AHUs - supply air temperature not tracking setpoint55% of AHUs
Terminal units - zone temperature sensor fault46%
Terminal units - abnormal zone temperature46%
AHUs with at least one fault on any given day40%
Terminal units with at least one fault on any given day30%

Common and persistent AHU faults: supply air temperature and static pressure setpoints, the economiser sequence, the supply air sensor, coil valves and airflow sensors.

Faults as reported by FDD tools in 317 North American buildings - a guide to where to look, not a UK survey. Source: Crowe et al. 2023 (LBNL) - fault detection data from 317 buildings, 3,660 AHUs and 53,865 terminal units. Flags: North American, Faults as reported by FDD tools.

Why delivery quality matters

One of the most useful findings for anyone pricing service is buried in the delivery-route analysis. Commissioning delivered through US utility programmes returned a median of about 5%. Monitoring-based programmes returned 9%. Properly scoped work done outside those programmes returned 14%. One programme that capped investigation budgets at $0.10 per square foot saved only about 3% of electricity.

The lesson is not that cheap work is worthless, but that the time spent investigating is what finds the savings. That is an evidence-based argument against the lowest service bid - and for scoping investigation time properly in a proposal. The LBNL authors also warn that published case studies are biased towards the best projects, which is one more reason to quote medians rather than standout results.

Savings decay without attention

Do savings last? The evidence is thinner than anyone would like. Texas A&M followed ten commissioned buildings over about eight years and found savings mostly held - but most of the losses came from mechanical failures in two of the ten (ESL-TR-07-01-01). The usual causes of decay identified in the literature are operators disabling optimised sequences, sensors and actuators failing, and staff turnover.

No robust year-on-year drift rate exists, so it is better not to quote one. The fair statement is simply that savings decay without attention - which is exactly what a service contract with monitoring is there to catch. The BMS service visit that pays for itself sets out a visit pattern built around that idea.

The honest UK gap

Almost every headline figure above is North American. The UK evidence is real but mostly qualitative. Innovate UK's Building Performance Evaluation programme, covering 50 non-domestic buildings, found controls to be a recurring problem - over-complicated systems, BMSs not integrated with other controls, and systems not commissioned properly before handover (Innovate UK 2016). The government's Building Energy Efficiency Survey modelled a 39% technical abatement potential in non-domestic buildings, with energy management, lighting and controls making up more than half of it (BEES) - again a model, not a measurement.

A claim that "90% of building controls are inadequate" circulates in the trade press, credited to the Carbon Trust. It could not be traced to a primary source and is best left alone.

No UK measured-savings dataset comparable to LBNL's exists. That is a weakness in the evidence - and an opening for integrators who measure, because a before-and-after result from a real UK building is worth more to the next client than any borrowed percentage.

How to talk about savings

A form of words that holds up:

That last line is the important one. Proving savings explains how to measure and report, and Fault detection on a BMS turns the fault list above into rules. For sector-specific angles, see BMS before heat pumps, Optimisation for landlords and Timetables, labs and cost per tonne.

Related tools and training: the BMS health check, energy savings estimator, CUSUM savings tracker and degree-day tools are in tecTools, and the Servicing and optimising an existing BMS course is in tecTraining.

Published evidence mentioned: Crowe et al. 2020 and Crowe et al. 2023 (LBNL, US); US DOE Smart Energy Analytics Campaign; Katipamula et al. 2021 and Fernandez et al. 2017 (PNNL, US, the latter modelled); Texas A&M ESL-TR-07-01-01 (US); Innovate UK Building Performance Evaluation (UK, 2016); Building Energy Efficiency Survey (UK, modelled). Results vary by building and none of these figures is a promise for any one site.

More articles from 3EC

3EC logo