We supply IT equipment and we also write software on it, which gives us an unusual vantage point. We see what gets bought, and eighteen months later we see what people are actually complaining about. The gap between those two things is remarkably consistent.
Here is what we tell clients before they sign a purchase order.
Specify by role, not by an average
The most expensive mistake in fleet purchasing is buying one machine for everybody. It is administratively tidy and it is always wrong in two directions at once.
The finance officer who lives in a browser and a spreadsheet has been sold a discrete GPU she will never wake up. The developer compiling a project forty times a day has been given 8 GB of RAM to keep the average down, and is now losing twenty minutes a day to swap, every day, for three years.
Three tiers cover nearly every business we work with:
- Standard. Browser, documents, email, video calls. 16 GB RAM, a 256 GB to 512 GB NVMe drive, integrated graphics, a genuinely good screen. The screen matters more than the processor for this person and it is where budget builds usually cut.
- Heavy. Developers, analysts, anyone running local databases or containers. 32 GB minimum (the single highest-return line item on the whole order) and 1 TB of fast storage.
- Specialist. Video, 3D, ML training. GPU-led, specified per person against the actual software they use, never against a category.
The cost difference between the tiered order and the flat one is usually small. The difference in daily friction is not.
RAM and storage are the whole game
If you take one thing away: for ordinary business work, RAM and SSD speed determine how fast the machine feels far more than the processor does.
A modest CPU with 32 GB and a fast NVMe drive will feel quicker in daily use than a top-tier CPU with 8 GB and a SATA drive, because what people experience as "slow" is almost always waiting on memory pressure or disk, not on compute. Retailers market the processor because it is the number on the sticker. Ignore the sticker.
Corollary: check whether the RAM is soldered before you buy. A machine you can upgrade in two years for a modest sum is a fundamentally different asset from one where the only upgrade path is a new machine. This detail is buried in the specification sheet and it is one of the highest-leverage things to check.
Cost the three-year total, not the sticker
Compare a $600 laptop against an $850 one and the cheap one looks obviously right. Then add the things that do not appear on the quote:
- Warranty length, and critically, whether it can be honoured locally. A three-year international warranty that requires shipping the unit out of the country is, in practice, not a warranty.
- Battery replaceability, because in a load-shedding environment the battery is not a convenience feature. It is uptime, and it is the component that degrades first.
- Whether replacement parts exist in the local market at all. Some brands you can get a keyboard or a charger for within a day. Others mean a two-week wait and a courier.
- Resale or redeployment value at end of life.
- The productivity cost of the downtime while it is being repaired.
Run that arithmetic and the $850 machine is frequently cheaper over three years, sometimes substantially. We have watched clients buy the cheap fleet, replace a third of it inside two years, and spend more than the good fleet would have cost, while their staff spent that time on worse machines.
The things people forget entirely
Docking and displays. A second monitor is one of the cheapest measurable productivity improvements available, and it is almost never in the original budget.
Chargers. Buy spares up front. A team sharing chargers because two went missing is a slow, silly, permanent tax.
Backup, on the day of deployment. Not later. A laptop is a device that will eventually be stolen, dropped, or drowned in a cup of tea, and the question is only whether that event costs you a machine or a machine plus everything on it.
Power protection. In an environment with unstable supply, a surge protector on every desktop and a UPS on anything that must not lose state is not optional spending. It is insurance priced at a fraction of what it protects.
Buy for the second year
The single question we ask that reframes most purchasing conversations: what will this person be doing in eighteen months?
Businesses grow into their software. The spreadsheet becomes a database. The single-location operation gets a branch. The designer who was doing static layouts starts editing video. Hardware bought precisely against today's requirement is obsolete exactly when the business is least able to afford disruption.
Specify with a little headroom. It is cheaper than replacing early, and far cheaper than a team quietly working around machines that are no longer up to the job.