Graduated Tech IndexOperational health ratings for graduated hardware brands

Methodology · algorithm version 1.9.0

How GTI grades a brand

A graduated brand raised demand through crowdfunding or a pre-order, early-access or reservation programme, or launched directly without one, and has since shipped product to paying customers. GTI asks one question of each: is the company operating like it intends to support what it sold?

Principles

  1. Algorithmic. Grades are computed from public operational signals only. There is no manual grade override.
  2. Independent of commerce. Commissions never affect ratings or ordering. Affiliate and commercial data is never an input to scoring, ranking or which brands are listed. See Disclosure.
  3. No credit for missing data. Every check ends in one of three states. Measured: we have the fact, including a confirmed “no” (for example, no SDK), which scores 0. Not applicable: the check does not apply (for example, a companion app for a robot that has none) and is left out. Unknown: we could not find the fact in public sources; it also scores 0, is labelled “Unknown” on the brand page, and lowers coverage. A brand with too much unknown is not graded.
  4. Polite collection. Official APIs first, plain HTTP for public pages, robots.txt respected, an identifying User-Agent, caching. No stealth or anti-bot evasion.

The three pillars

The Stability Score (0–100) combines three pillars. Each pillar is built from one or more observable signals, and a signal that does not apply to a brand (for example a companion app for a product without one) is left out rather than penalised.

  • Firmware & Software Health. Whether the product is still being maintained: how recently software last changed (releases or code updates on the brand's official GitHub repositories, iOS and Android app updates), on how many days in the past 12 months there was software activity, and how recently the companion app was updated. Full marks need a change within the last 7 days, activity on 24 or more days a year (about twice a month) and an app update within 14 days; nothing in 90 days scores 0.
  • Warranty & Support. The length of the standard warranty the manufacturer offers (statutory guarantees required by law in some regions are not counted), and the App Store rating of the brand's companion app. The rating counts only with at least 50 ratings; it rates the app, not the hardware or the support desk.
  • Ecosystem Maturity. The footprint around the product: whether an SDK (software development kit) or API is available, so developers can build apps and integrations for the device, and the range of first-party accessories and replacement parts.

Pillar weights

CategorySoftwareWarranty & SupportEcosystemWhy
AR / XR glasses40%35%25%A display that depends on firmware and a companion app to stay compatible with phones and laptops, and fragile hardware worn daily, so software and warranty lead; developer apps and accessories add value but most buyers never use them.
3D printers35%30%35%A machine with wearing parts that is only as useful as the parts, upgrades and printer software around it, so ecosystem counts as much as software; warranty matters but wear parts are replaced, not claimed.
Humanoid robots40%20%40%Bought mostly by developers and labs, who need ongoing software and an SDK to build on; warranty terms are usually set per quote and few have a rated consumer app, so that pillar says less.

Data sources

  • Official public APIs: GitHub (releases and code updates on the brand's official repositories) and the Apple iTunes Lookup API (app update dates and App Store ratings).
  • Brave Search API: the date of the latest Android app update shown in search results. Google Play is never fetched directly.
  • Researched facts (warranty length, SDK, accessories), each with a source URL and date. Where a brand publishes no warranty terms, the warranty stated by an authorised reseller is used and cited.

Grades

Scores map to a letter scale from A+ (strongest) downwards. Only grades of C+ and above are published, and each is shown with the score and the data coverage behind it.

Coverage

Data last updated: 2026-11-17. All brands are refreshed together in each data run.

Coverage is the share of the scoring model that had fresh data for a brand. A check with no data scores 0 points, so a small public footprint lowers both coverage and the grade. A brand without enough public data for a grade is not shown.

Comparison

The buyer-matchmaking block (“Which graduated brand fits your workflow?”) is generated from the data on the page: a brand gets a reason only where it leads the brands compared on a measure taken from cited public data (for example the most recently updated companion app, the most stable software releases in 12 months, the longest stated warranty, or the largest disclosed funding). Brands are listed by grade, then score; brands still being collected follow alphabetically. The “Check pricing & bundles” button goes to an affiliate link when one is active and otherwise to the official store; affiliate data plays no part in which brands appear, in what order, or in which reasons are produced. See the disclosure.

Limitations

  • Android app update dates are read from search results, not from an official store API, so they can be missing or approximate; Android-only apps can lose coverage.
  • Software activity is only visible where a brand publishes releases or code on GitHub or ships a companion app. A brand that updates firmware privately scores lower on Firmware & Software Health.
  • Warranty length is a single figure per brand. Where terms differ by product or region, the standard figure for the main product is used.
  • Grades are opinions derived from this methodology and cited, dated evidence. They are versioned, so a change to the algorithm is visible as such.

Contact

Questions about the methodology: GTI@vinsights.club.