Rippa · Machinery
A dealer storefront engineered from 57,494 owner messages
Rippa is a line of Chinese mini-excavators. Buyers are cautious — it's a $6,700+ machine and the internet is full of clones and half-truths. The site answers the two questions that actually close the sale — "is this the right machine for my job?" and "what will owning it really be like?" — the first with a proper catalogue, the second with a knowledge base written from a live owners' community, not from marketing.
Inside the product
The gap
A brochure site answers neither question a cautious buyer actually has. Anyone can list horsepower; almost no dealer can show eight months of what breaks, what to check and what the machine really costs to run — in the owners' own words.
The insight
A real owners' community was the moat. A points programme can be copied; an audience network can't — and neither can eight months of genuine field experience, if it could be turned into a searchable, credible knowledge base rather than copywriter fiction.
A product tour
A conversion-focused catalogue
Five models (R06 → R22) with full spec sheets and a story-driven flagship page, honest "from $" pricing, in-stock vs. pre-order badges, a 12-type attachments catalogue, parts and consumables, and one-tap Telegram/WhatsApp/phone contact — so a buyer self-qualifies before ever messaging the dealer.
A community-sourced knowledge base
The full export of the owners' Telegram — 57,494 messages, 383 owners over eight months — parsed, threaded and scored; the top threads rewritten into 60 articles across 8 categories, with every quote still credited to its real author, date and message id. Editorial, not invented.
Photos matched to topics
A script matches owner photos to each article by scanning the surrounding conversation — a photo's message plus its reply parent and children — then optimises them to responsive WebP at 800/1600/2400 px.
Every buyer question, answered first
The catalogue and knowledge base were shaped around real pre-purchase anxieties: which machine to pick, registration and licensing, how to haul it, running costs and part numbers, how to price your own work, and how to spot a fake.
Under the hood
Trilingual with a transliteration engine
Russian and Uzbek-Latin are authored by hand; Uzbek-Cyrillic is generated on the fly by a longest-match transliterator (digraphs, apostrophe letters, protected brand names). One language, maintained for free.
Static pre-render for SEO
After the Vite build, Puppeteer walks every route in all three languages and freezes it to static HTML, so Google and Telegram previews get fully-rendered pages with per-language JSON-LD and sitemaps.
A real data pipeline
Eight Python scripts (~950 lines) parse 58 HTML export files into a structured message stream, rebuild and score reply threads by usefulness, select candidates, and match photos — plus image optimisation to responsive WebP.
Stack & channels
Services behind this
The outcome
A storefront that both sells the machine and earns the buyer's trust — a due-diligence tool built from an original data pipeline turning 57,494 community messages into 60 owner-grade articles. Shipped to production.
Start a project
Tell us what's missing.
One form, a real reply within 24 hours. Or book a call directly.
Other ways to reach us
- Emailhello@thnkers.com
- LinkedIn@thnkers
- Book a callcalendly.com/thnkers