A critical spares inventory is a stock of replacement parts held close enough to restore a failed device inside its recovery objective. Support contracts promise four-hour response or next business day, but those numbers describe an obligation, not a physical fact. The physical fact is that a replacement part has to travel to the failed device.
Most of the time the two align. The times they do not are the times that matter: a regional event where every customer is calling at once, a part that has gone end-of-life without anyone noticing, a site that is four hours from the nearest depot on a good day and unreachable on a bad one. A spares inventory is the answer to the question of what happens when the contract does not perform.
Which Parts Deserve a Spare?
Parts whose failure takes down a service and whose replacement lead time is long; most others do not, because spares tie up capital and they expire. Stocking broadly is expensive and produces a shelf of hardware that no longer matches what is installed. The inventory has to be selective, and the selection criteria are worth being explicit about.
Start from failure impact rather than failure likelihood. A component whose failure takes down a service is a candidate regardless of how reliable it is. A component with a redundant partner is a much weaker candidate, because the redundancy already bought the time that a spare would buy.
Then consider replacement time. A part available from three distributors with overnight shipping does not need to sit on a shelf. A part with a twelve-week factory lead time does, and the length of that lead time is the single strongest argument for holding stock.
Single points of failure are the core of the list: the core switch with no pair, the storage controller in a simplex configuration, the transfer switch, the environmental monitoring unit that nobody thinks about until the room gets hot. So is anything where the failure cascades — a power distribution component whose failure takes a rack, a network element whose failure isolates a site.
Cost enters last. A cheap part with a long lead time and a high impact is the easiest decision in the inventory. An expensive part with a short lead time and a redundant partner is the hardest, and usually the answer is no.
Where Should Spares Be Kept?
Close enough to reach the failed device inside the recovery objective. A spare in a central warehouse serves many sites efficiently and serves none of them quickly. A spare in the room serves one site instantly and is idle the rest of the time. Most organizations need both, distributed according to how fast the recovery has to be.
The test is whether the part can reach the failed device inside the recovery objective that your contingency plan sets, under the conditions where it will be needed. That last clause matters. A regional warehouse two hours away is two hours away in normal weather. During the ice storm that caused the outage, it is not.
Remote and hard-to-reach sites justify local stock at a level that would be wasteful elsewhere, because the transport time dominates everything else. Sites with a DR partner nearby can sometimes share, which halves the holding cost if the failure modes are genuinely independent.
Vendor-managed forward stocking is an alternative worth pricing. Several manufacturers will position inventory in a depot near a customer site under contract, which converts a capital purchase into a service and shifts the obsolescence risk. It works well for standard parts and poorly for anything unusual.
Spares Rot
An inventory that is not maintained becomes a shelf of equipment that cannot be used, and the discovery happens during the incident.
Firmware drifts. A spare switch that has been in a box for two years runs a version that will not join the current fabric, and updating it under pressure requires a management connection, a firmware file, and a procedure that nobody has rehearsed. Spares should be updated on the same cycle as production, which means they need to be powered up periodically rather than stored.
Configuration is the same problem. A spare controller with factory defaults needs the production configuration applied before it is useful. Your IT team keeping current configuration copies accessible, and not on the system being replaced, is what makes a spare deployable in minutes instead of hours.
Batteries self-discharge and degrade whether or not they are used. Optics and cables get borrowed. The most common cause of a missing spare is that someone took it for a project and the shelf was never reconciled.
Compatibility drifts as production changes. A hardware refresh that replaces half the fleet makes half the spares irrelevant, and the spares inventory is rarely part of the refresh plan. It should be — the review belongs at every significant change, not at an annual interval.
How Should You Track Spares?
Track them like assets, not supplies, in the same kind of hardware inventory that CIS Control 1 calls for. The spares list needs to be a real record: part number, serial number, quantity, location, firmware level, date acquired, date last verified, and the systems it supports.
That last field is what makes the inventory useful during an incident. Someone with a failed component needs to know in seconds whether a spare exists and where it is, and a list organized by part number does not answer that. A list organized by the systems it protects does.
Reconcile physically. A quarterly count against the record finds the borrowed parts, the units that were consumed and never replenished, and the entries that were removed from production two refreshes ago. Replenishment should be automatic when a spare is used, and it usually is not, because the incident ends and the urgency ends with it.
Test Before You Need It
A spare that has never been installed is an assumption. Deploying one during a maintenance window — actually swapping it into production, confirming it works, and either leaving it in place or returning it to the shelf — converts the assumption into a fact and exercises the procedure at the same time.
It also finds the things that only appear on contact: the mounting rail that is a different generation, the license tied to a serial number, the cable that shipped with the original and not with the spare. Those discoveries are cheap on a Tuesday afternoon and expensive at two in the morning.