Eleven configurations in a fleet of 200 is not variety, it is four separate costs you are paying without a line item: dock types to stock, spares that fit nothing, support knowledge nobody has, and resale as several small mixed lots instead of one clean batch.
TL;DR
- Fragmentation is almost always caused by writing the specification as a minimum rather than as exact figures.
- Two configurations is the right number for most fleets: one standard, one for heavy workloads.
- Do not replace machines to achieve consistency. Apply the standard to new purchases and let the refresh cycle do it.
- Name a model plus two alternates per market, because a single named model will be out of stock somewhere.
- The substitution reporting route matters more than the standard. Unreported substitutions are how you got to eleven.
- Count your distinct configurations today. That number is the honest measure of the current arrangement.
How fleets fragment
Three causes, and the first accounts for most of it.
The specification was written as a minimum. A standard saying at least 16GB of memory invites whoever is ordering to meet it with whatever is available, and over two years that produces five or six configurations without anybody breaking a rule.
A named model went out of stock and somebody substituted sensibly. Entirely reasonable, and the substitution was never reported because there was no route to report it.
Exceptions were granted individually and never fed back. Each one was justified. Collectively they became the fleet.
None of those is a discipline failure. They are all a reporting failure, which is why a stricter standard does not fix fragmentation and a substitution route does.
What fragmentation costs
| Dock and adapter types | One per configuration family | Mis-pairing is routine, and it surfaces when you equip an office |
| Spares that fit nothing | A spare for one configuration is not a spare | You hold more machines to cover the same risk |
| Support knowledge | Eleven machines to know rather than two | Diagnosis is slower and more often wrong |
| Resale value | Several small mixed lots | A worse rate per unit than one clean batch of the same total |
| Imaging and testing | Each configuration needs validating | Mostly invisible until something breaks on one model only |
The resale row is the one that produces a number somebody notices, because it arrives as a quote rather than as an ongoing cost. It is also the point at which fragmentation is too late to fix for those machines.
Collapsing to two
Two configurations, defined by workload rather than by seniority.
A standard configuration for work that is a browser, a document editor, communication tools and video calls, which is most people. Specify exact memory and storage figures, not minimums, and set the memory generously because it is what determines whether the machine is still pleasant in year 4.
A higher configuration for work that compiles, renders, runs containers or virtual machines, or handles large media. Defined by what runs on it, which means a junior engineer qualifies and a senior manager doing browser work does not.
Then resist a third. The pressure for one comes from a handful of genuine edge cases, and those belong in an exceptions route with a log. A third configuration reliably becomes the one people negotiate for, and you are back to fragmenting from a higher starting point.
A worked example
A company with 190 people across 12 markets wrote a standard naming one model and one configuration. It held in 4 markets.
In 3 more the model was available with a 5-week lead time against 3 days at home. In 2 it came only in a configuration one step up. In the remaining 3 whoever was ordering locally substituted something close and told nobody, because there was nowhere to tell.
Eighteen months later the register held 11 distinct configurations and nobody had broken the standard. The standard had described an outcome supply could not produce, so every market resolved it locally and the resolutions never came back. The fix was a model table with two alternates per market and a one-line substitution reporting rule.
Getting from eleven to two
What to do:
- Count your distinct configurations from the register today, before deciding anything.
- Define two configurations with exact figures rather than minimums, named for the workload.
- Build a model table: one row per market with the named model, two alternates and the supplier.
- Write the substitution rule in one line, naming who may substitute and where they record it.
- Apply it to new purchases only and let the refresh cycle retire the rest over a few years.
- Review the model table quarterly and the configuration definitions annually, since they change at different speeds.
The fifth point is the one companies get wrong in the other direction. Replacing working machines to achieve consistency is expensive, disruptive and reads as a reallocation exercise. Doing nothing except not adding a twelfth configuration gets you there in about three years at no cost.
Disclosure: RemoAsset is owned by the same people who publish PeopleOpsHQ. It is relevant here only because the delivered configuration is recorded against the asset at the order rather than being assumed from the policy, which is how unreported substitutions become visible. It publishes no price and requires a demo, it is much weaker for hardware it did not supply, and it is not a certified disposal vendor. The options are compared in our laptop procurement platform comparison.
Final thoughts
Fragmentation is not caused by people ignoring standards. It is caused by standards that describe an outcome supply cannot deliver, with no route to report the sensible substitutions that follow.
Write two configurations with exact figures, name a model and two alternates per market, add a one-line substitution reporting rule, and apply it only to new purchases. Count the configurations you have now so you can tell whether it worked, and review the model table every quarter because that is the part that actually changes.
Frequently asked questions
How many laptop configurations should a company have?
Two, defined by workload rather than by seniority. A standard configuration for work that is a browser, a document editor and video calls, which covers most people, and a higher one for work that compiles, renders, runs containers or handles large media. Specify both with exact memory and storage figures rather than minimums, since a minimum invites whoever is ordering to meet it with whatever is in stock and produces five or six configurations within two years.
What does a fragmented laptop fleet actually cost?
Four things that rarely appear as line items. Dock and adapter types, one family per configuration, with mis-pairing becoming routine and surfacing when you equip an office. Spares that fit nothing, so you hold more machines to cover the same risk. Support knowledge, since diagnosing eleven machines is slower and more often wrong than diagnosing two. And resale value, because several small mixed lots fetch a worse rate per unit than one clean batch of the same total.
Should you replace machines to standardise a fleet?
No. Replacing working hardware to achieve consistency is expensive, disruptive and will be read as a reallocation exercise rather than an improvement. Apply the new standard to new purchases only and let the refresh cycle retire the rest, which takes about three years and costs nothing beyond the discipline of not adding another configuration. The one exception is a machine that genuinely cannot do its user’s work, which is a capability decision rather than a standardisation one.
Why does a single named model not work across markets?
Because availability and lead time diverge. The same configuration can be in stock in one market and five weeks out in another, available only one step up in a third, or sold through a channel that does not support automated enrolment in a fourth. Whoever orders locally then faces a delay nobody will accept or a substitution nobody will notice, and they substitute. Name a model plus two alternates per market in a table you review quarterly.
What is the single most important part of a fleet standard?
The substitution reporting rule, not the specification. Fragmentation happens because sensible substitutions are made under time pressure and never reported, since there is nowhere to report them. One line naming who may substitute, what they may substitute to, and where they record it prevents the outcome that stricter specifications cannot. Without it the register describes what the policy said rather than what was delivered, and the divergence is invisible until a resale quote.
How do you know whether you have a fragmentation problem?
Count the distinct model and configuration combinations in your register and compare that number to the number your standard describes. That single figure is the honest measure of how well the current arrangement works, and it is usually higher than anybody expects. If the register records the policy’s configuration rather than the delivered one, which is common, check a sample of machines against their rows instead, because the register may be understating the problem.
