- Home
- Planned shutdowns
Taking a building’s plant offline
Switching a unit off takes thirty seconds. Everything either side of it is where jobs go wrong, and almost all of it can be agreed a fortnight in advance.
Four questions decide every shutdown, and they are the same four whether the job is an hour or a week.
Who loses what, and for how long. Not “the building” — which units serve which spaces, and which of those spaces has somebody in it who will notice. In a multi-tenant building this is frequently unknown until somebody traces it, and tracing it is a survey rather than a guess.
What has to be isolated, and by whom. Electrical isolation, gas isolation and controls are three separate things and they are not always ours to do. On a building with a management system, taking a unit off without telling the controls means the system spends the day trying to start it.
Who has to be told, and in writing. Tenants, the managing agent, security, anybody whose access is affected, and anybody who has to be present. The written part matters because “I mentioned it to somebody” is not a record.
And what the restart looks like. More jobs are judged on the restart than on the work. A building that comes back on and is comfortable by opening time was planned. One that is still recovering at eleven was not.
Pick why the system is coming off and what the building holds. Everything below is written out in full.
If a building holds stock, samples, reagents, food or medicines that depend on temperature, somebody owns that risk and it is not the mechanical contractor. We are not qualified to say what your product tolerates, we have no visibility of your obligations, and we will not sign a method statement that implies we do.
What we can tell you, and will, in writing: which spaces lose conditioning, at what point in the day, for how long, what the space is likely to do while it is off (based on the building rather than on hope), and how long it takes to recover afterwards. That is genuinely useful and it is the limit of what we know.
What you need alongside it. Somebody who owns the product decision — a quality manager, a pharmacist, a food safety adviser, whoever it is in your organisation. Independent monitoring during the work, logged, rather than a thermometer somebody remembers to look at. A written decision about whether stock moves, and where to. And an agreed stop point: a temperature or a time at which we put the system back regardless of how far the work has got.
That last one matters. Agree it in advance, in writing, and make it ours to act on. A crew that has to ring somebody for permission to stop is a crew that does not stop.
The most predictable of the five and the one most often underestimated, because the mechanical work is the smallest part of the day.
The sequence that actually governs it: isolation, disconnection, lifting the old unit off, curb adaptation if the new one does not match, setting and sealing, reconnecting gas, power and controls, and commissioning. The curb is the variable. A replacement that sits on the existing curb is a different job from one that needs an adaptor fabricated, and that is knowable from the survey rather than on the day.
The crane is the constraint. Where it can stand, what it has to lift over, whether the car park has to be closed, and whether the street needs anything. It also fixes the date far more rigidly than anything else, which is why the rest of the plan is built around it.
And the roof is somebody’s. Removing and setting a unit disturbs flashing and sometimes the membrane. Whether your roofer attends, or inspects afterwards, should be agreed before the day rather than argued about after a rainstorm.
The job where the HVAC contractor is not in charge and everything still lands on the HVAC, which makes early involvement worth a great deal.
Units usually have to be lifted and reset, which means isolation, disconnection, temporary support or removal, and reconnection and recommissioning afterwards. That is two mobilisations, not one, and it wants pricing as two.
Curbs are the decision nobody makes in time. A re-roof changes the height of the deck, which changes how the curb relates to it. Whether curbs are raised, replaced or adapted is a question for the roofer and us together, and it is much cheaper to answer on a drawing than on a roof.
And the building is usually occupied. A phased re-roof means phased shutdowns, which means the tracing exercise — which unit serves which space — has to be done properly before the first phase, not discovered during the third.
The one thing worth insisting on: get the roofer and us in the same conversation before the programme is fixed. It costs an hour and it routinely saves a week.
Straightforward in principle and the one most likely to damage equipment, because plant that is simply cut off and cut back on has a worse time than plant that is shut down properly.
Before: systems brought down in order rather than by pulling a main. Controls told, so a management system is not fighting an absence. Anything with a crankcase heater noted, because a compressor restarted cold after a long outage is a compressor at risk.
After: systems brought back up in order, not all at once — simultaneous starts on a building full of units is a genuine inrush problem and occasionally trips the supply that was just restored. Then verified: each unit actually running, not merely powered.
The part people forget: a brownout or a dip is worse than a clean outage for motors, and an electrical upgrade day frequently involves several. If the work will produce dips rather than a clean off, tell us, because the correct answer is then to have the plant genuinely off rather than riding through them.
The shutdown that is rarely called a shutdown, which is exactly why it causes trouble.
Dust is the first problem. Running a system in a space being demolished or cut pulls construction dust straight through the coil and into the ductwork, and the tenant next door gets it. Either the unit comes off and the openings are sealed, or filtration is upgraded temporarily and somebody owns changing it. Deciding that after the drywall sanding starts is too late.
Shared units are the second. In a great many small commercial buildings one unit serves more than one tenancy, or a return path crosses a demising wall that is about to be moved. That has to be established before the fit-out programme is set, because it turns one tenant’s work into another tenant’s shutdown.
And the handover is the third. A fit-out changes the load, the layout and often the duct. Whether anybody rebalances afterwards is usually nobody’s line item, and it is the reason a newly fitted suite is uncomfortable from the first week.
Then the sequence is compressed but it is still a sequence, and the temptation to skip it is what turns a bad day into a bad fortnight.
First, establish what is actually off rather than what is reported off. In an unplanned failure the reported symptom and the affected area often do not match, and half an hour of tracing prevents a wasted mobilisation.
Second, decide the holding position before the repair. Temporary cooling, moving people, restricting a space, shutting a section — those decisions belong to you and they are easier with an honest estimate of how long the repair will take, including the parts.
Third, be told the truth about parts. On older rooftop equipment the honest answer is sometimes that the part is days away, and that changes every decision you make. A contractor who says “we’ll get it going” without knowing is not helping you.
And afterwards, write it down. An unplanned failure is the best information you will ever get about that machine. It belongs in the record, with what failed and what it took, because that is what makes the next one planned.
The simplest case, and the one where the honest answer is usually about tolerance rather than risk.
An office without conditioning for part of a day is uncomfortable rather than dangerous, and it recovers. What decides whether it is a problem is timing and communication: the same shutdown is a non-event on a Tuesday morning in March and a complaint on a Thursday afternoon in September.
What we would agree in advance: which spaces, what hours, and whether the building can be occupied while the work happens. And whether ventilation is affected as well as cooling — a sealed office with no outside air gets stuffy long before it gets hot, and that is the complaint people actually make.
If anyone in the building has a specific requirement, tell us. We will work around it if we can and say so if we cannot.
A kitchen changes a shutdown completely, and mostly because of ventilation rather than cooling.
Extract and make-up air are a pair. Taking a make-up air unit off while a hood keeps running puts the whole building into negative pressure — doors that will not shut, and in a building with atmospheric gas appliances a genuine concern about how they vent. Taking the extract off means the kitchen cannot operate at all. Which of those you are choosing needs stating out loud.
Gas appliances and gas isolation. If gas is being isolated for our work, somebody has to know which appliances that affects and who relights them. That is frequently not us.
And the practical one: kitchens run on a service timetable that does not bend. Work happens between services or overnight, and a plan that assumes a Tuesday afternoon is free has not spoken to the head chef.
We will not advise on food safety. What we will do is say clearly what the ventilation will and will not be doing, and when.
The room that does not care what day it is, and the one most often left off the shutdown plan because it has no people in it.
It runs all year, frequently on its own unit, and in small buildings that unit is sometimes the only thing on a circuit somebody is about to isolate. Establish what serves it before anything is switched off, not after somebody notices the temperature alarm.
A small sealed room heats fast. Faster than people expect, because everything in it is converting electricity into heat continuously. If its cooling is coming off for any meaningful period, the realistic options are temporary cooling, propping the door and moving air, or a controlled shutdown of the equipment — and the third one is an IT decision, not ours.
Who owns it. In a lot of buildings the server room cooling was installed by somebody else entirely, is maintained by nobody, and is discovered during a shutdown. If that describes yours, the survey is worth doing before the shutdown is planned.
See the note above, which is the position and does not move.
The practical addition is about sequencing. Where stock is involved, the shutdown should be planned backwards from what the space can do rather than forwards from what the work needs. That frequently means overnight or weekend work, a shorter off-period split across two visits, or temporary cooling — all of which cost more, and all of which are cheaper than the alternative.
It also means the survey comes a long way in advance, because the answer sometimes is that the work cannot be done while the stock is in the building.
The next step
Get the curb surveyed before anything is ordered, and get the crane position agreed before a date is offered to anybody. Those two decide the shape of the whole day.
The next step
Get the roofer and us in the same conversation before the programme is fixed. One hour, and it settles curb heights, phasing and who is responsible for the flashing.
The next step
Tell us whether the day produces a clean outage or a series of dips. They want opposite handling, and dips are worse for the equipment than being properly off.
The next step
Establish which units and return paths cross the demising walls before the programme is set, and decide now who owns rebalancing at handover. Both are cheap today and expensive in week six.
The next step
Establish what is genuinely off before anybody mobilises, and get an honest answer about parts lead time. Those two let you make the holding decisions that are actually yours to make.
We will not take responsibility for what a temperature excursion does to your product. Stated at the top of this page, in the refusals table on the home page, and in every method statement we sign. It is not a negotiating position.
We will not promise uptime. No percentage, no guarantee of continuous operation. Plant fails, and a contractor promising otherwise is describing a hope. What we can offer is a record good enough that most failures stop being surprises.
We will not start a shutdown without a written stop point where one is warranted. Agreed in advance, and ours to act on without ringing anybody.
And we will tell you when the work should not happen yet. If the survey is not done, if the tracing is unclear, if nobody has told the tenants — we would rather move the date than start and find out.
Plan it a fortnight out
A survey, the tracing done properly, and a written shutdown plan: what comes off, when, who is told, what the restart looks like and where the stop point is. It is the cheapest part of any job of this kind.