Retrofit & economics

The savings were in the controls, and so was the complaint

Why a controls retrofit routinely delivers a larger saving than the luminaire swap and a larger number of complaints, what false-off actually costs, how to size daylight zones that work, standby power on a large estate, and the commissioning settings that decide whether any of it survives the first month.

Written by Yilian Zhikong Technology (Shenzhen) Co., Ltd. · Last reviewed August 2026

Two retrofits on the same building. The first replaces every fitting with a more efficient one and cuts consumption by a predictable fraction. The second leaves the fittings alone and adds presence detection and daylight regulation, and cuts it by more. The second also generates the complaints, gets partially disabled by facilities within six weeks, and ends up delivering a third of what was modelled. Both outcomes are normal, and the difference between them is almost entirely commissioning.

Where the saving actually comes from

A luminaire swap reduces the power drawn while the lights are on. Controls reduce the hours they are on at all, and in most commercial buildings the hours are the larger term. An open-plan office lit from 07:00 to 20:00 for the benefit of a cleaning shift and two early arrivals is burning most of its energy on unoccupied space, and no efficacy improvement addresses that.

The saving splits into three mechanisms that are worth estimating separately, because they behave differently and stack multiplicatively rather than additively:

• Absence or presence detection — the hours where the space is empty and the lights were previously on. Largest in cellular offices, meeting rooms, WCs, stores and plant rooms; smallest in continuously occupied open plan.
• Daylight regulation — the reduction in output while useful daylight is available. Concentrated in the first few metres from the glazing and worth very little beyond that, which is why the zoning matters more than the sensor.
• Time scheduling and curfew — the blunt instrument, and often the one that delivers most on the first day for the least money.

False-off is the whole risk

Every complaint about a controls retrofit reduces to the lights going off while somebody was still there. It is worth being precise about why this matters more than the inconvenience suggests: the first time it happens, the occupant loses trust in the system; the second time, they find the override; the third time, facilities are asked to disable the sensors, and the project's saving disappears while its cost stays.

The causes are mundane and mostly designed in rather than faulty.

1. Detection pattern versus the actual task. Passive infrared responds to movement across its field, not towards it, and responds far better to a person walking than to a person typing. A sensor sited for coverage of a circulation route will not reliably see someone reading at a desk under it.
2. Partitions and furniture installed after the sensor layout. High storage turns one detection zone into two, and one of them has no sensor.
3. Time-out set from the default. The out-of-the-box value is chosen to demonstrate the product, not to suit an office. Too short and it becomes a complaint; too long and the saving evaporates. It is a per-space decision.
4. Zone too large. One sensor controlling a large open area means the last person present holds the whole zone on, which is safe and saves nothing, or that a sparsely occupied area goes dark in places, which is a complaint.

Daylight zoning that survives contact with a window

Daylight regulation fails in a different way: it works, and then somebody complains that the light level changes visibly, or that the row nearest the window is dim on a bright day while the second row is not.

Three rules cover most of it. Zone parallel to the glazing, in bands, because the daylight gradient runs perpendicular to it — a zone that spans from the window to the core will always be wrong at one end. Site the sensor to see a representative surface rather than the sky or a single bright desk, and be aware that a repositioned desk or a new partition changes what it sees. And set a slow fade rate: the eye detects change, not level, and a regulation system that adjusts in visible steps will be noticed and disliked even when every level it produces is correct.

Standby power, which nobody models

Every sensor, every control interface and every driver in a controlled installation draws power when the lights are off. Individually it is fractions of a watt to a couple of watts. Across a large estate with several thousand control points and a long unoccupied period, it becomes a line item that offsets a measurable part of the saving.

Ask for the standby figure per control device and per driver, multiply by the count and by the unoccupied hours, and put it in the model as a negative. On schemes with an aggressive out-of-hours strategy it has occasionally been large enough to change the zoning decision — fewer, larger zones with fewer control points beating a fine-grained scheme that saves more in theory.

Five things to specify before the sensors are ordered

1. A zoning drawing showing detection zones and daylight bands against the actual furniture layout, not the empty shell.
2. Time-out, fade rate and daylight set point specified per space type, in a schedule, rather than left to the commissioning engineer's defaults.
3. A manual override in every space, and a documented statement of what it does and how long it lasts. An override that occupants trust is what stops them from asking for the system to be disabled.
4. Standby power declared per device, with the estate total in the energy model.
5. A four-week settling period after handover with a scheduled revisit to adjust settings, priced into the contract. Every controls project needs this and almost none of them have it, which is the single largest reason the modelled saving and the delivered saving diverge.