Transparency
How We Score
RadarTrek scores are designed to help you make faster, better decisions — not to fill space or drive affiliate clicks. This page explains exactly how we arrive at every score, what we have verified ourselves, and what we have sourced from third parties.
What we have not done (yet)
We have not independently run every speed test, set up every test hosting server, or personally benchmarked every AI model from scratch. Where we have not run our own tests, scores are aggregated from publicly available benchmarks, audits, and independent reviews — and we say so clearly in each category's methodology note.
This is our starting point, not our ceiling. As RadarTrek grows, we are building a testing infrastructure to replace aggregated scores with independently verified ones. The path: VPN speed tests first (one VPS, one script), then hosting benchmarks, then structured AI output evaluation.
General principles
Scores are 0–100, higher is better
Every dimension score is normalised to a 0–100 scale. A score of 80+ is strong, 60–79 is good, 40–59 is average, below 40 is weak. Scores are relative to the tools in the category — a 70 in one category is not the same as a 70 in another.
Overall score is a weighted average
The overall score combines all dimension scores using category-specific weights. Dimensions with weight 2 count twice as much as weight 1. For VPNs, speed and privacy have weight 2 because they matter more to most buyers than, say, device support limit (weight 1).
No tool pays to appear or be ranked
Affiliate links are present on this site — when you sign up through our links we may earn a commission. Affiliate relationships have zero influence on scores or rankings. Tools are ranked purely by their weighted score across our research-based dimensions.
Scores are point-in-time snapshots
Pricing changes. Features launch and disappear. Audit reports are published. We aim to review scores every 3–6 months and update when significant changes occur. The date a category was last reviewed is noted in each methodology section below.
Suggest a correction
If you find a factual error — wrong pricing, a missing audit report, a score that does not match published benchmarks — email us at hello@radartrek.com with a source link. We investigate and update within 48 hours.
Per-category research sources
VPNs
VPN scores are aggregated from published independent speed tests (Cloudwards, PCMag, TechRadar), audit reports (Cure53, KPMG), official pricing pages, and simultaneous connection limits. Speed scores reflect average download speed retention vs baseline across multiple server locations. Audit scores reflect whether a formal no-logs audit has been published, who conducted it, and when.
Web Hosting
Hosting scores are aggregated from Review Signal's WordPress Hosting Benchmarks, GTmetrix performance tests on identically configured test sites, independent uptime monitoring data, published support response time studies, and official pricing pages. Performance scores reflect Time to First Byte (TTFB) from multiple global locations. Uptime scores reflect 12-month monitoring averages from third-party tools.
AI Tools
AI tool scores are based on published model benchmarks (MMLU, HumanEval, HELM), pricing as listed on official websites, context window sizes from official documentation, independent multimodal capability tests, and API documentation quality assessments. Output Quality scores reflect aggregate performance across writing, reasoning, and coding benchmarks. Scores reflect capabilities as of mid-2026.
Indie SaaS Stack
SaaS stack scores are based on free tier limits from official pricing pages, developer experience ratings from the Indie Hackers community, Stack Overflow Developer Survey results, official SLA documentation, documentation quality assessments, and community activity (GitHub stars, Discord size, Stack Overflow question volume). Scores reflect the perspective of a solo or small-team developer building a web product.
Code Editors
Code editor scores are based on published developer surveys (Stack Overflow, JetBrains State of Developer Ecosystem), community ratings, official pricing and free tier limits, independent multi-file editing capability tests on standardised codebases, and documentation quality. AI Quality scores reflect benchmark performance on HumanEval and similar coding benchmarks where published. Multi-file editing scores reflect observed capability depth from community evaluations.
Website Builders
Website builder scores are based on official pricing pages, template gallery assessments, Google Lighthouse performance scores on published test sites, published SEO capability comparisons, and e-commerce feature lists from official documentation. Portability scores reflect the difficulty of migrating content away from each platform based on documented export tools and community migration reports.
How scores will improve over time
VPNs
Automated speed tests from a VPS in three regions (UK, US, Asia) against each provider's closest server. TTFB and download speed retention vs unprotected baseline.
Web Hosting
GTmetrix and WebPageTest runs on identically configured WordPress test sites hosted on each provider. TTFB from five global locations, monthly for 12 months.
AI Tools
Standardised prompt battery across writing, reasoning, and coding tasks. Blind scoring of outputs against a rubric. Repeated quarterly as models update.
Code Editors
Timed completion of a standardised feature implementation task (add authentication to a Next.js app). Measured from prompt to working code without manual editing.
Affiliate disclosure
Some links on RadarTrek are affiliate links. When you sign up or purchase through these links, we may receive a commission at no additional cost to you. Affiliate relationships are disclosed in the footer of every page and do not influence scores, rankings, or editorial decisions. We recommend tools based on score, not commission rate.
Questions or corrections: hello@radartrek.com