How the Mars Score Works: The Full Formula, Explained

vsMars is a transparent product comparison platform. The Mars Score is its open scoring formula — a composite computed from a product's published specifications using per-category weights that anyone can inspect and recompute. This article is the canonical reference for how it works. The full dataset behind every score — every product, every spec, every result — is published as open data (CC BY 4.0).

The formula in one paragraph

Every category (smartphones, headphones, GPUs…) defines a set of scored specs grouped by theme (display, battery, performance…), each group carrying a weight. Each spec value is normalized to 0–1 against a published range using one of three methods, averaged within its group, and the groups are combined as a weighted average into a raw 0–100 score. That raw score is then mapped through a fixed, monotonic display rescale into the 35–95 presentation range you see on product pages. No editor adjusts it; no advertiser can move it.

Step by step

1. Collect specs. Every product carries a complete spec sheet sourced from manufacturer data and retailer APIs (see our methodology for sources).
2. Normalize. For each scored spec, the value maps onto 0–1 against the category's published min–max range. Three methods exist: minmax (higher is better — battery capacity), minmax_inverse (lower is better — weight, response time), and log_minmax (diminishing returns — resolution-class specs). A 7,000 mAh battery in a range spanning 3,000–8,000 mAh normalizes to 0.8.
3. Average within groups, then weight across groups. Each spec group's normalized values are averaged, and group results combine by the category's weight table — smartphones weight battery and display heavily; monitors weight panel quality.
4. Guard on confidence. The formula tracks what fraction of required specs the product actually declared. Below 40% coverage we refuse to show a score at all — a score computed from a half-empty spec sheet is worse than no score.
5. Rescale for display. The raw score maps through 35 + (raw/100)^0.85 × 60. Anchor points: raw 0 → 35, raw 50 → 67, raw 70 → 79, raw 90 → 89, raw 100 → 95. The mapping is monotonic, so relative ranking between any two products is preserved exactly — it only shifts the absolute numbers into the range buyers intuitively expect, without inflating budget products.

What the score is honestly not

• It is not a review. It measures what the spec sheet declares, weighted by what typically matters in the category. It cannot see haptics quality, software polish, or how a camera actually renders skin tones — that's what Mars Labs hands-on reports are for.
• It is not a star rating. We deliberately do not publish the Mars Score as an AggregateRating in structured data — search engines treat self-assigned ratings as self-serving. Ratings on vsMars come only from real user reviews.
• It is not universal across categories. A 78 in smartphones and a 78 in e-readers are computed against different ranges. Compare scores within a category; across categories, compare the spec sheets.

Why ranges, not observed min/max

Normalizing against the live catalog's observed extremes would make every product's score shift whenever any other product enters the catalog. Published fixed ranges keep scores stable, reproducible, and honest: you can take a product's spec sheet, the range table on the methodology page, and this article, and arrive at the same number we show.

Recomputation

Scores are recomputed from specs at build time — there is no cached editorial number that can drift from the data. Change a spec value or a weight, and every affected score, comparison verdict, and ranking updates in the same deploy. The weights themselves are versioned; when a category's weight table changes, the version number in the dataset changes with it.

Check our work

The entire dataset is downloadable as JSON with full spec sheets or flat CSV. If you recompute a score and get a different answer, report it — a transparent formula is only worth something if people can hold it to account.