Media · software reviews
Software review site: weighted scoring, rated evidence and quality gates before every deploy
An English-language review site for chatbot platforms: hands-on reviews scored on a weighted 17-dimension rubric, an evidence rating for every claim, interactive calculators and a build that fails on broken links.
- Industry
- Media · software reviews
- Technologies
- Next.js 16 · React 19 · TypeScript
Challenge
Small and mid-sized businesses choosing a chatbot platform mostly find review pages that look mass-produced and never say what was actually tested. The aim was a review site that earns trust by showing its work: how each score is made, how solid the evidence behind each claim is, and what the prices mean for a real monthly budget. The same rules had to hold on every page as the site grew.
Solution
We built the site on Next.js with MDX content. Each review is written after hands-on testing and scored on a 17-dimension weighted rubric, and the score sidebar shows every weight. An evidence box rates each claim on a five-step scale, from tested hands-on to not verified, and a fixed formula turns that into an overall evidence level. A value score compares price with the cheapest comparable platform. Around 20 calculators turn pricing into monthly costs, and the build checks that numbers quoted in the text still match what the calculators return.
How it connects
- 01 Question research
- 02 Hands-on testing
- 03 Weighted scoring
- 04 Quality gate
- 05 Link-checked build
- 06 Verified deploy
What we delivered
- Long-form reviews with a weighted rubric and a visible score breakdown
- Evidence box with a five-step rating per claim and an overall evidence level
- Value-for-money score relative to the cheapest comparable platform
- Best-of lists, head-to-head comparisons, alternatives pages, a glossary, channel guides and a blog
- About 20 interactive calculators, with build-time checks that quoted numbers match their output
- Filterable review hub
- Disclosed partner links, and a newsletter sign-up with bot protection and rate limiting
Functionality
- A broken internal link fails the build
- Every page passes a QA checklist, voice rules and a quality-rater-style review before it ships
- Integrity rules: no invented measurements, no fake bylines, and free-plan testing never presented as paid-plan testing
- Structured data for Review, SoftwareApplication, ItemList, FAQPage and BreadcrumbList
- Every deploy is compared with the live sitemap and the list of expected URLs, and redeployed if pages are missing
- Reviews are written by hand; only glossary, blog and list pages are drafted with automation
Integrations
- Serverless PostgreSQL through Prisma
- Newsletter sign-up with bot protection
- Partner link tracking
- Scheduled jobs on Vercel
Technologies
Result
Live and in operation, with daily deploys checked against the live sitemap.
Why is every deploy checked against the live site?
Once, the site kept serving an old build for two weeks while the homepage still returned 200, so uptime checks saw nothing wrong. Since then every deploy compares the pages on disk with the live sitemap, its dates and every URL that should exist, and redeploys on its own if something is missing. A green deploy log no longer counts as proof that new pages are live. The same principle runs through our editorial pipeline for a Portuguese-language publication.
Discuss your project
Tell us about your project