A weighted-criteria approach to buying a laptop

Laptop reviews on YouTube are not bad. They are optimized for a different problem than yours. A reviewer needs to say something distinctive about 50 laptops a year to keep an audience. They can’t tell you whether the M3 MacBook Air is right for you — they can only tell you how it compares to the last 49 laptops they held. Your buying problem is more specific than that, and it benefits from a framework instead of a reviewer.

Here is a weighted-criteria approach to laptops that produces a defensible pick in about 15 minutes. The choose-laptop template in Phronesis runs this structure for you.

Why reviews mislead

A reviewer is evaluating against the category. You are evaluating against your workflow. Three concrete failure modes:

The benchmark arms race. Reviewers test on synthetic workloads (Cinebench, Geekbench, PugetBench) because those numbers are comparable across laptops. Your actual workflow is probably one of: heavy browser tabs + slack + zoom, IDE + docker + browser, Lightroom + Photoshop, or CapCut/Final Cut. Synthetic benchmarks are a proxy for your workflow, not a measurement of it. The proxy breaks down on anything memory-bandwidth-sensitive or thermally-bursty.

The delight overweighting. Reviewers spend 45 minutes with a laptop and weight their first impressions heavily. Your first impressions will fade in a week; your irritations — the trackpad click that is too shallow, the port you wish was USB-C, the weight in your bag — will compound for three years. Review-based decisions overweight the former.

The new-features trap. New hardware gets reviewed more than mature hardware. A two-year-old ThinkPad is a rational pick for many developers, but it does not get reviewed because it is not new. Review-dominated buying systematically underweights mature, known-good options.

Starting with your workflow, not the laptops

Before you look at any product page, write down what you actually do with your laptop. Be concrete:

• Hours per week on each primary application
• Battery dependence (coffee-shop hours per week vs always-plugged)
• Travel pattern (backpack vs desk, flights per month)
• Peripheral setup (single screen? dual external? dock?)
• Thermal tolerance (do you care if your legs get warm?)
• Keyboard preference (key travel, layout, trackpad vs mouse)

This takes ten minutes. It is where most buyers skip the real work. The workflow doc is the input to your criteria — without it, you are choosing between “good laptops in the abstract” instead of “the best laptop for your actual life.”

The weighted criteria

A reasonable weighted model for a developer / knowledge worker:

• Sustained performance under load — 20%
• Battery life at realistic workload — 15%
• Keyboard + trackpad quality — 15%
• Display quality (color, brightness, refresh) — 10%
• Weight + build + portability — 10%
• Port selection + docking behavior — 10%
• Repairability + upgrade path — 5%
• Ecosystem (macOS vs Windows vs Linux fit) — 10%
• Price — 5%

Weights should sum to 100%. You will notice price is only 5% — that’s intentional. Unless you are on a hard budget cap, the cost difference between laptops in your shortlist is usually $300–$700, and spread over 3 years of ownership the per-day cost is ~$0.50. Weighting price heavily makes you optimize for a dimension that matters almost nothing compared to daily annoyances.

For a creator (video editor, photographer), reweight: Display up to 20%, Sustained performance up to 25%, Battery down to 10%.

For a traveler, reweight: Weight + portability up to 20%, Battery up to 25%, Performance down.

For a student on a budget: Price up to 25%, Ecosystem up to 15%, others down proportionally.

The fact that you have to change the weights by persona is the point. A reviewer giving a universal answer cannot be right for all three personas simultaneously.

Running the comparison

Pick 3–4 candidates. Typical shortlists in 2026:

• Developer, macOS-friendly: MacBook Pro 14 M4, MacBook Air 15 M4, Framework 13 with AMD Ryzen AI
• Developer, Linux-friendly: Framework 13, ThinkPad X1 Carbon Gen 13, Dell XPS 14
• Creator, video-heavy: MacBook Pro 16 M4 Max, ASUS ProArt Studiobook, MacBook Pro 14 M4 Pro
• Traveler, budget: MacBook Air 13 M3, Lenovo Yoga Slim 7, Framework 13 base config

Drop your shortlist and weights into Phronesis — three blind analyst models score each on each dimension, return a ranked output with agreement level and per-dimension scores. What matters most in the output:

• Where did the analysts agree? Those dimensions are stable across priors.
• Where did they disagree? Those are typically subjective dimensions (keyboard feel, ecosystem fit). Your own preference settles those.
• What was the margin? If the winner won by 0.2 points and the confidence is 55%, your shortlist is too homogenous. Add a genuinely different option (e.g., if all three are ultrabooks, add a workstation).

The underweighted variables

Three things nobody weights enough:

Thermal behavior under sustained load. Every modern laptop boosts aggressively for 30 seconds and then thermally throttles. If your workflow includes 15-minute compile cycles, large video exports, or ML training, sustained performance matters far more than peak. Intel 14th/15th-gen mobile chips run 15–25% below peak under sustained load; Apple Silicon holds closer to 90%. This is buried in benchmark footnotes and it should be on the front page.

Trackpad quality. You use the trackpad thousands of times a day. A bad one is a permanent tax on your attention. Apple’s trackpad remains meaningfully better than the Windows competition, which has closed but not matched. For developers who live in the terminal this matters less; for designers who live in UI tools, this is close to a first-class criterion.

Fan noise at idle + light load. Some laptops spin fans audibly during a Zoom call while your Mac neighbor sits in silence. If you take a lot of calls, this is a quality-of-life dimension. It is almost never in reviews.

Repairability. If you plan to keep the laptop 4+ years, repair-ability matters. Framework is the outlier. Apple is now mediocre-but-improved. Dell XPS is getting worse. This is a long-horizon variable that most reviews downweight because their audience upgrades every 2 years.

The ecosystem question

For many buyers, ecosystem is the actual tiebreaker. If you already live in iMessage, FaceTime, AirDrop, and Continuity, switching to Windows breaks your cross-device workflow in painful ways. If you already live in WSL2, PowerToys, and Windows-specific enterprise software, switching to Mac breaks yours.

Ecosystem is not a tiebreaker that can be argued with specs. It’s a tiebreaker that deserves an explicit weight in the model. Give it 10–15% unless you genuinely don’t care (rare).

What about the “AI PC” marketing?

2025 and 2026 have been the years of “AI PC” marketing — laptops with NPUs marketed as future-proofing for AI workloads. The honest assessment: for most knowledge workers, the NPU is irrelevant today and probably will be for another year. The actually consequential AI capabilities are running cloud models (bandwidth + latency matter, not your NPU) or running a small model locally (a 40+ TOPS NPU is the floor, but most 2024-era laptops already clear it). Don’t pay a premium for NPU marketing in 2026 unless your specific workflow uses local models daily.

The contrarian move

Most buying advice points you at the newest thing. The contrarian move is to buy the second-newest flagship. A year-old M3 MacBook Pro is close to the M4 in real workflow performance and is ~20% cheaper on the used market. A one-generation-back ThinkPad X1 Carbon has the same keyboard, the same ports, and a tested-not-early-adopter firmware.

The exception is the Mac transition off Intel. There, buying old is buying wrong, because the architectural upgrade is meaningful. Otherwise, the marginal upgrade between generations is often smaller than the secondary-market discount.

The decision protocol

1. 10 minutes: Write your workflow doc. Be specific about hours, peripherals, and travel.
2. 5 minutes: Pick weights. Persona-based starting points above, adjusted for your workflow doc.
3. 5 minutes: Build a shortlist of 3–4 candidates (the contrarian move: include one last-gen flagship).
4. 2 minutes: Run Phronesis using the choose-laptop template.
5. 10 minutes: Read the per-dimension scores, not just the winner. Look at the dimensions where the analysts disagreed — those are where your subjective preference matters.
6. Outside the tool: Go to a store and physically touch the top two. Trackpad, keyboard, weight in a bag, screen in your lighting. If the #2 feels right and #1 feels wrong, trust the touch test — Phronesis can’t feel the trackpad.

That’s about 30 minutes from “I need a laptop” to “I have a defensible pick.” Compare that to the 6 hours of review-watching most buyers do — same outcome quality, 12× faster, and you end up with a written justification you can show to the person paying for it.

The broader point

Laptop buying is a specific case of a general problem: a consumer category with more review content than buyers can process, and almost no guidance on how to filter the reviews against your workflow. The fix is the same fix that works for CRMs, frameworks, hires, and strategy: a written criteria model with weights, a blind evaluation, and a commitment to read disagreements as signal.

Don’t trust the reviewers. Trust your criteria.