How We Score and Compare Products: A Worked Example

vsMars publishes two kinds of judgement about products: a Mars Score for each product, and a comparison — which product wins on which spec — for any two products in the same category. Neither is written by hand. Both come from deterministic code that runs over the published spec sheets, and both can be reproduced by anyone with the spec data and the rules on this page.

This guide explains that code in plain language, and then walks a real product through every step so you can see the arithmetic. If you want the formula on a single page, read How the Mars Score works. For data sources and every category's range table, see our methodology. This page is the bridge between the two: a worked example.

Principles first

Four rules shape everything below.

1. Specs in, score out. The score is computed only from a product's spec sheet. No editor can raise or lower a product's score directly, and no partner can pay to change it. To change a score, you have to change a spec value or the category's published weights — and both are visible.
2. Fixed ranges, not live ones. Each scored spec is compared against a fixed range defined per category, not against the best and worst products currently in the catalog. That keeps a product's score stable when other products are added.
3. Missing data is missing, not zero. If a manufacturer does not publish a figure, it is left out of the calculation rather than scored as the worst possible value.
4. Compare within a category. Each category has its own specs, ranges and weights. A 78 for an earbud and a 78 for a laptop are not comparable numbers.

Step 1: Every category defines what is scored

Each category in our catalog has a spec configuration: the list of specs we track, grouped into themes, and a flag on each spec saying whether it counts toward the score. Many specs we display — codecs, Bluetooth version, water rating — are useful for comparison but are not part of the score.

For wireless earbuds there are four groups, with these weights:

Note the connectivity row. The earbuds connectivity group currently contains Bluetooth version, codecs, multipoint and in-ear detection, and none of those is marked as counting toward the score. A group with no scored values is skipped, and the remaining weights are re-normalized — so in practice, for earbuds, audio counts for 45/85 (about 53%), battery for 30/85 (about 35%) and design for 10/85 (about 12%) of the result. We state that here because it is a real property of the current formula, and it explains why, for example, multipoint support does not move an earbud's score even though it appears in every comparison.

Step 2: Each spec is normalized to 0–1

Raw specs are in different units — millimetres, hours, grams — so each scored spec is first converted to a number between 0 and 1 using its category range. The code supports these methods:

• minmax — higher is better. A value at the bottom of the range scores 0, at the top 1, linearly in between. Values outside the range are clamped to 0 or 1.
• minmax_inverse — lower is better (weight, response time). The same, reversed.
• log_minmax — higher is better, with diminishing returns: the scale is logarithmic, so doubling a small value counts for more than doubling a large one.
• boolean — yes is 1, no is 0.
• enum_rank — for ordered categories, each step up the list counts evenly between 0 and 1.

For earbuds the ranges are: driver size 6–16 mm (minmax), battery in the buds 3–12 hours (minmax), weight per bud 3–12 g (minmax_inverse), ANC (boolean).

Clamping has visible consequences. The Master & Dynamic MW09 lists 16 hours of battery in the buds and the Anker Liberty 5 lists 12. Both are at or above the 12-hour ceiling, so both normalize to 1.0 on that spec: past the ceiling, extra hours stop adding score.

Step 3: Average within groups, weight across groups

Within each group, the normalized values are averaged equally, then multiplied by 100. The groups are then combined as a weighted average using the category's weights (re-normalized over the groups that have data, as above). The result is the raw score, from 0 to 100, rounded to one decimal place.

Step 4: The confidence guard

At the same time, the code counts how many of the category's scored specs the product actually has a value for. That fraction is the confidence. If it is below 40%, we do not show a Mars Score at all — a number computed from a mostly empty spec sheet would look precise while meaning little. Products above the threshold but below 100% show their score with a note that it is based on partial data.

Step 5: The display rescale

Raw scores are mapped to the 35–95 range shown on the site with one fixed formula:

display = 35 + (raw / 100)^0.85 × 60

So raw 50 displays as about 67, raw 70 as about 79 and raw 90 as about 89. The mapping is monotonic: if product A has a higher raw score than product B, it also has a higher displayed score. It changes how the numbers look, never the order. The displayed value is rounded to one decimal place.

Worked example: Samsung Galaxy Buds 3 Pro

The Galaxy Buds 3 Pro spec sheet in our catalog lists, among other things: 10.5 mm drivers, ANC, 6 hours of battery in the buds and 5.4 g per bud. Those are its four scored specs.

Normalize:

• Driver size: (10.5 − 6) ÷ (16 − 6) = 0.45
• ANC: yes = 1
• Battery in the buds: (6 − 3) ÷ (12 − 3) = 0.333
• Weight per bud (lower is better): 1 − (5.4 − 3) ÷ (12 − 3) = 0.733

Group scores:

• Audio: (0.45 + 1) ÷ 2 × 100 = 72.5
• Battery: 0.333 × 100 = 33.3
• Design: 0.733 × 100 = 73.3
• Connectivity: no scored specs — skipped

Weighted average: (72.5 × 0.45 + 33.3 × 0.30 + 73.3 × 0.10) ÷ 0.85 = (32.63 + 10.00 + 7.33) ÷ 0.85 ≈ 58.8 raw.

Confidence: all four scored specs present = 100%.

Rescale: 35 + (0.588)^0.85 × 60 ≈ 73.2.

That 73.2 is exactly the Mars Score shown on the Galaxy Buds 3 Pro page. The same steps for the AirPods Pro 3 — 11 mm drivers, ANC, 8 hours, 5.3 g — give group scores of 75.0, 55.6 and 74.4, a raw score of about 68.1, and a displayed 78.3.

The breakdown also tells you why they differ. The audio and design groups are nearly identical. Almost the entire 5.1-point gap comes from the battery group: two extra hours in the buds is worth roughly 22 points within that group, and the group carries about 35% of the effective weight.

How comparison pages decide who wins

A comparison page — for example AirPods Pro 3 vs Galaxy Buds 3 Pro — does not use the Mars Score to pick winners spec by spec. It uses a separate engine that compares the two spec sheets directly, over the category's list of headline comparison specs. For earbuds those are: battery in the buds, battery with the case, ANC, driver size, weight per bud, water rating, spatial audio and multipoint.

For each spec, the engine decides a winner by data type:

• Numbers: higher wins, unless the spec is marked lower-is-better (like weight). Equal values tie.
• Yes/no features: having the feature wins. Both or neither ties.
• Ordered categories: the value further up the published order wins — for water resistance, the order runs IPX0, IPX2, IPX4, IPX5, IPX7, IP54, IP55, IP57, IP67.
• Versions: compared part by part, so 5.4 beats 5.3.
• If either product is missing the value, that spec is left out rather than awarded to the other side.

Each win is then described with its size — as a percentage for specs like battery, or an absolute difference for specs like weight — and the advantages are sorted so that the biggest differences appear first.

For the AirPods Pro 3 against the Galaxy Buds 3 Pro, that produces:

• AirPods Pro 3 win: battery in the buds (8 h vs 6 h, 33% more), battery with the case (30 h vs 26 h, 15% more), driver size (11 mm vs 10.5 mm, 5% more), and weight per bud (0.1 g less).
• Galaxy Buds 3 Pro win: multipoint.
• Tie: ANC, spatial audio, IP57 water resistance.

Two things are worth noticing. First, a "win" is a statement about the spec sheet, not about importance: a 0.1 g weight difference is a technical win that nobody will feel, and the page shows the size of every difference so readers can judge that for themselves. Second, the comparison can surface things the score ignores — multipoint and water rating do not count toward the earbuds score, but they do appear in every earbuds comparison.

The written verdict on each comparison page is generated from the same data: it names each side's largest advantages, quotes the exact figures, and states both Mars Scores. It contains no claims that are not in the spec sheets.

One URL per comparison

Comparison URLs list the two products in alphabetical order of their URL slugs — AirPods Pro 3 vs Galaxy Buds 3 Pro, never the reverse — and the reverse order permanently redirects to it. That way every pair has exactly one page.

What the score deliberately does not do

• It is not a review. It measures what the spec sheet declares. It cannot hear tuning, feel comfort or measure how strong noise cancellation actually is. That is what our hands-on labs are for, and our editorial guides say explicitly when the data cannot answer a question.
• It does not include price. A product's Mars Score is the same whether it costs $99 or $349. Our guides discuss value separately, using the list prices in our catalog.
• It is not a star rating. We do not publish Mars Scores as star ratings in search-engine structured data. Ratings presented that way come only from real user reviews.
• It does not know your phone. Codec support, platform features and ecosystem integration often matter more than any single spec — see Galaxy Buds vs AirPods for a case where they decide the answer.

Known limitations, stated plainly

A formula is only as good as its inputs and its choices, and we would rather list the weak points than have readers discover them:

• Specs are proxies. Driver size is a rough proxy for audio capability, not a measurement of sound quality. Weight is a proxy for comfort.
• Ranges are judgement calls. Choosing 3–12 hours as the battery range for earbuds is a decision. A product beyond the ceiling gets no extra credit, and a different range would produce different scores — though the ranking within the range would not change.
• Unscored specs still matter. As described above, earbuds connectivity currently contributes nothing to the score. Comparison pages compensate by showing those specs side by side, but the score alone does not reflect them.
• Manufacturer figures are claims. Battery life is generally quoted by manufacturers under their own test conditions, often with ANC off. We record what is published; real-world figures are usually lower.

When we change a weight or range, the category's weights carry a version number, and every affected score, ranking and comparison is recomputed from the specs in the same release — there is no stored editorial score that could drift from the data.

Check our work

Every product's spec sheet is on its product page, and the full dataset — products, specs and scores — is published as open data. The ranges and weights for every category are listed in our methodology. With those, a spreadsheet and the five steps above, you should arrive at the same number we show. If you do not, we would like to hear about it — please report it.

For a buyer's view of the same data, see best wireless earbuds in 2026 and best earbuds for small ears.