The two questions that decide everything

Every resilience conversation reduces to these, and they are worth answering separately because they lead to different equipment.

How long can you be without power before it hurts? Seconds for a data room. Minutes for a shop. Perhaps hours for a workshop. That answer determines whether you need something instant or something that starts.
How long might the outage last? Most are under an hour. A cable fault or storm damage can be a day or more. That determines whether a battery is enough or something has to generate.

Why both matter

A generator takes ten to thirty seconds to start and transfer. That is fine for lighting and fine for most machinery. It is not fine for anything that cannot lose power for a moment — servers, controls, process plant mid-cycle.

Conversely a UPS that holds for fifteen minutes is no use in a six hour outage. Most sites that need one need both, and the useful conversation is about where the boundary between them sits.

Putting a number on it

Work through these and you will have a figure worth making decisions against.

Cost	How to estimate it
Lost output	Hourly revenue or production value, times realistic outage hours per year.
Idle staff	People being paid to wait. Often the largest single line and the most overlooked.
Restart time	Rarely instant. Process plant, ovens, chillers and IT all take time to come back — sometimes longer than the outage.
Spoilage	Chilled and frozen stock, work in progress, anything mid-process that cannot be salvaged.
Contractual	Service level penalties, missed delivery windows, and the customer who does not come back.
Data and equipment	Corrupted databases, damaged drives and control systems that need an engineer to reset.

What we would look at on your site

Your actual load, measured. Not the sum of the nameplate ratings, which is almost always double what you really draw.
Starting currents. Motors, compressors and chillers can draw six or seven times running current for a moment. That is what sizes a generator, and it is the commonest reason a set that "should" be big enough is not.
What is genuinely critical. Splitting essential from non-essential loads is usually the single biggest cost saving available — you back up a third of the site instead of all of it.
Your existing switchgear. Where a transfer switch can go, what space exists, and what the earthing arrangement will allow.
Fuel and access. How long you can run before someone has to deliver, and whether they can get to you in the conditions that caused the outage.

What a plan should contain

Equipment is only part of it. A resilience plan that has never been read is not a plan.

Which loads are backed up, and which are deliberately not
What happens automatically, and what needs a person
Who that person is, and who it is when they are on holiday
How long you can run unaided, and who you call for fuel
A test schedule — monthly off-load, annually under load

The test nobody does

A generator that starts every month on no load will still fail when it is asked to take the site. Off-load running is not a test; it glazes bores and proves almost nothing.

At least annually, a standby system should be proved under real load — either your site load or a load bank. It is the only way to find the problems before the outage does.

What we provide

Power quality metering and demand analysis
A written resilience assessment with the costs above quantified for your site
Specification and installation of UPS, generation and transfer equipment
Load bank testing and commissioning
Ongoing maintenance and annual proving

If the assessment concludes you do not need much, we will say so. A business that loses two hours a year to outages and can absorb them does not need a standby generator, and we would rather tell you that than sell you one.

Work out what an outage costs you. Then decide.