Laptop sizing for IT teams

Buying laptops by job title and gut feel produces a fleet that is too big for some people and too small for others. Measured use fixes that, if the measuring is done fairly.

Guide. For IT and procurement teams sizing a fleet.

Decide what you are sizing for

A fleet is bought for roles, not for individuals. Start by listing the roles that share a configuration, such as field engineers, analysts and office staff. Each role needs a question you can answer from data: does memory pressure reach the warning band during their normal day, and how often is the CPU or GPU the limit instead?

Write the question down before any measurement starts. A question chosen after the data arrives tends to fit whatever the data shows.

Who to measure

Measure several people per role, and include people who run the heaviest version of the work, not only the most typical ones. One volunteer is an anecdote. Two or three from different teams make a pattern you can check. Ask people to keep working as they normally would, and do not tell them which applications to open, because that would measure a day nobody actually has.

Tell people what is captured and where it is stored before you start. Measurement that feels like surveillance produces bad data, because people change how they work when they feel watched.

How long to measure

One to two weeks of normal use. A short window can miss the heavy days, and the first day often differs from the rest of the week, because people are settling in. Two weeks covers a full working cycle for most roles, including the busy days that generate complaints.

SizeMyLaptop records a capture every 30 minutes while the machine is awake, so the sample is steady across the period rather than concentrated on the days someone remembered to check.

Peaks matter more than averages

An average memory figure is the wrong summary for a sizing decision. A laptop that averages comfortably but spends two hours a day in critical pressure will generate complaints, and the average will never show why. Look at how often each band was reached, how long the sustained stretches lasted, and which applications were open at those times.

Report the share of captures in each band for each role, and keep the worst sustained period next to it. Those two numbers say more than any single mean.

Set the thresholds before the data arrives

Agree on what counts as too small before you look. For example: a role is undersized if pressure reaches the critical band during more than a set share of its working captures. Pick the share, the band and the window in advance, write them down, and apply them to every role the same way. This stops the thresholds drifting toward the answer someone wanted.

Keep the raw reports. Anyone who challenges a decision should be able to see the numbers behind it.

Keep the data local

Captures can show which applications people use and when. Treat them as personal data. Keep the reports on the machines they came from, or in a restricted folder, and share only the aggregated result that the decision needs. SizeMyLaptop writes everything locally and uploads nothing, which makes that handling simpler to explain.

Common questions

How many machines should I measure?

Enough to cover each role you buy for. Sample several people per role, and do not treat one volunteer as the whole group.

Why not use averages?

Averages hide the moments that cause complaints. Look at how often memory pressure reached the warning and critical bands, and at the sustained stretches.