The standard

The Deskmark

A Deskmark is our verdict on where a product stands in its class — and the one thing about it no company can buy. It grades pay-blind standing, computed from a published algorithm on sourced, dated data. Not a lab test. Not sponsored. Not for sale.

7,344 listings graded, in 701 of our 782 device and software categories · 1,704 more On the Record · 9,048 in all, not counting 1 product an authority has told people to stop using, which holds no Deskmark.

The ladder

Platinum The single best-graded entrant in its class, and only in a class with at least 8 graded. 535 one per class
Gold The top fifth of its graded class, rounded up, apart from a Platinum leader. 1,214 17% of graded
Silver The next band below Gold, reaching to the middle of its graded class (always at least one place). 2,087 28% of graded
Bronze Graded, and ranked below Silver. 3,508 48% of graded
On the Record Sourced and dated, but not graded: a device whose record carries fewer than half of its class's comparable specs, or whose class has no comparable spec or fewer than 5 products; or software whose record lists no feature or integration. 1,704 the gate

How a Deskmark is earned

To carry any metal, a product must first be On the Record — every spec and price sourced to the public page it was captured from, the seller's or maker's own wherever they publish it, and dated; where a figure can't be sourced, we publish a null, never a guess. Its tier is then its standing in its category, from the same public algorithm that powers our rankings. A device is graded on value for money, spec strength, and how recently the model was released. Software is graded on value for money, the breadth of its listed features and integrations, and how openly the vendor publishes its price. Platinum is reserved for the single category leader, so the apex stays rare by construction.

Devices — a 0–100 score from three published inputs:

Value for money 45%

Spec strength 35%

How recently the model was released 20%

Value for money is spec strength per dollar, with the price measured against the median of the class. Spec strength places each of the product's published numbers between the weakest and strongest in its class and averages them. It counts only comparable specs: ones where more (or less) is better and at least two products in the class have a number on record. A product's record must carry at least half of those for it to be graded, so a class with no comparable spec has no graded device. Recency of release compares the model's release year with the newest in its class, and that score halves for every two years behind. A product with no price or no release date is scored on the inputs it has; one an authority has told people to stop using is not graded.

Software is weighed in the same shape. A price list has no release date, but it does have a choice to publish — that takes recency's place:

Value for money 45%

Breadth of listed features and integrations 35%

How openly the price is published 20%

Value for money is the number of distinct features and integrations listed on the vendor's record, per dollar of its entry price, ranked within its class. The entry price is the lowest price the vendor lists for a paid plan, shown with its unit on the "Entry price" line of its record: a monthly price for nearly every vendor, and a yearly, one-time or per-return price where that is the lowest it lists. Breadth is that same count, ranked within the class; a vendor whose record lists neither stays On the Record. Openness is full when the entry price is a figure, a free plan's $0 included, or reads "usage-priced"; half when it reads "not priced monthly", as it does for a price in a currency or shape we cannot compare; and zero when it reads "quote-only". Features and integrations are counted as the pages we cite list them; we have not tested them.

A vendor whose entry price reads "quote-only" also scores zero on value, so it always ranks below a vendor in its class that lists as many features and integrations and is not quote-only. Any other vendor with no paid price to divide by (its entry price is a free plan's $0, or reads "usage-priced" or "not priced monthly") scores the middle of its class on value.

What a Deskmark is not

It is not a hands-on lab test. We are a sourcing and verification desk, not a testing bench — a Deskmark grades the standing of a product on its sourced record, the way a credit rating grades standing from disclosed data. And it is never for sale: no vendor can buy, rent or lobby for a tier, a score or inclusion, and asking alone moves nothing. A product's standing changes only when our reading of the public record changes (a vendor or a rival changes what it sells or publishes, or we read further or correct a reading), when its class gains or loses an entrant, or when the published, versioned formula changes. Every grade in a class is recomputed together, because each grade is relative to its class, and the formula never reads commission data.

Use it

Building an assistant or a storefront? The Deskmark and the sourced data behind it are queryable — use it in your AI assistant, take the open data, read the accuracy commitment it rests on, or report a correction (always free, always logged).