I use the rule of thumb that 10%, minimum 1, of active components should be spared.
This means that if there are ten or less of item A then you maintain one spare on site. Twenty or less then you keep two on site.
I don't bother sparing "inactive" components. Terminals, wire, mounting hardware, etc.
Lastly maintenance or "consumable" items, like fuses, batteries, etc. I usually go with 100% spares. i.e. I could replace every fuse once. Most likely these aren't big ticket items anyway.
Also as the designer, you should look to maximize commonality of parts. For example don't spec a 1.2KVA and a 1.5KVA drive or a 30W and 50W power supply. Just make both 1.5 KVA drives and both 50W power supplies. Any savings you might have will be eaten up by having to spare both.
This is what I would recommend. Now, they might be in for some sticker shock. At which time you can look at the lead time of the various items. Of course they need to compare this with cost of the equipment not working.
My typical sites use $1 per second for downtime. So, it doesn't take long to lose any savings by not carrying spares.
There are algorithms for estimating probabilities, risks, and costs associated with critical spares if you want to go down that route but you have to collect reliability information, lead times, downtime values, etc. to use them accurately which is why I think most people just use quick rules of thumb. I've done the calculations and generally they closely matched the recommended spare from the simple rule of thumb. Besides most of those calculation only really work well if you are deploying a fleet of machines.
The one thing I don't see very many people consider these days in obsolesce of the parts. One advantage of having a spare on hand is that if a device is obsoleted without direct replacement you have time to re-design before it breaks again (hopefully).