Messaging · B2B software research
WhatsApp Business platform ratings portal: plan-level scores and real monthly costs in two languages
A bilingual ratings portal for WhatsApp Business platforms in one Latin American market, with plan-level scores that show their sources and the real monthly cost of each platform in local currency.
- Client
- Client confidential
- Industry
- Messaging · B2B software research
- Year
- 2026
Challenge
WhatsApp Business platforms price in very different ways: per seat, per message package, with Meta's fees passed through at a markup, with or without a setup fee. Support in the local language, local payment methods and local invoicing vary widely, and Meta's own rates change on dated announcements. A buyer comparing list prices cannot see what a platform will actually cost per month.
Solution
We built a bilingual application with its methodology stored as data: weighted criteria, sub-criteria and scoring components over a registry of facts. Each platform is scored on the plan a buyer would actually use, with an evidence ladder and an upper bound that shows how much is still unverified. A cost engine prices a standard small-business scenario in local currency from the plan, the seats, dated Meta rates and the markup. Pros, cons and fit warnings come only from trigger rules over the data.
How it connects
- 01 Platform research
- 02 Fact registry
- 03 Rating and cost engines
- 04 Bilingual website
- 05 REST API and MCP
- 06 Editorial rules
What we delivered
- Bilingual app: English at the root, the second language under its own path, reciprocal hreflang pairs and localized 404 pages
- Methodology as data: weighted criteria, sub-criteria and scoring components tied to individual facts
- Scoring, total-cost-of-ownership and reputation engines writing to one committed ratings store
- REST API with an OpenAPI document
- Read-only MCP server over the same ratings store
- MDX articles in paired language versions, with numbers inserted only through data components
- Animated home-page hero: a WhatsApp conversation in which an AI agent takes a local instant payment, with a reduced-motion fallback
- Build gates, unit tests and end-to-end scenarios on desktop and mobile profiles
Functionality
- Plan-level scores with visible sources, an upper bound and the share still unverified
- Monthly cost in local currency for a standard small-business scenario on every platform card
- Cited strengths, weaknesses and fit warnings generated by trigger rules
- Publishing blocked unless each language version has at least three cons and one fit warning
- Coverage of the official WhatsApp API path, AI chatbots, CRM, local payment methods and local-language support
- Rankings, platform comparisons and recommendations
- Language switch that leads to the paired page in the other language
- MCP tools for search, platform cards, comparisons, pricing scenarios, integration guides and score explanations
Integrations
- Model Context Protocol server over HTTP, read-only
- REST API (/api/v1) with an OpenAPI document
- Dated Meta WhatsApp Business rates feeding the cost engine
- Review-platform snapshots feeding the reputation score
- llms.txt and JSON-LD structured data
- Vercel Web Analytics
Technologies
Result
Built and operated, live and indexed, with rated platform cards and a first series of paired articles in both languages, and more articles scheduled. Pages, API, MCP server and articles read the same ratings store, so a new Meta rate or a corrected price reaches all of them in one build.
One ratings store, four readers
The portal keeps its facts in git and compiles them to JSON at build time; there is no database. Scores, costs and reputation go into a single committed ratings store that the pages, the REST API, the MCP server and the articles all read. Articles receive numbers only through data components (score, price-from, setup fee, cost-of-ownership tables, Meta rates), so an article cannot quote a different figure from the platform card it links to.
Two languages, one site
English sits at the root and the second language under its own path. Every page has a reciprocal hreflang partner, the language switch lands on the paired page, and tests keep the two UI dictionaries key-identical. A spell-check for the second language runs with the other build gates: data sync, score verification, link audit, content gates, publish check, ranking preview and a confidentiality scan.
Discuss your project
Tell us about your project