Owned product3 review platforms · 25 sample records

Cross-Platform Customer Review Scraper

Data / Backend Engineer · Portfolio project · Cross-platform review extraction

A Python scraping toolkit that collects public customer reviews and app metadata from Trustpilot, Google Play, and the Apple App Store for customer research and sentiment analysis.

View source & evidence

Normalized review JSON · source URLs preserved.

How it works

Discover

Profile URLs + package/app IDs

Fetch

Patchright + platform endpoints

Extract

Reviews + app/business metadata

Normalize

Analysis-friendly JSON records

Deduplicate

Review IDs + fallback keys

Store

JSON outputs + trimmed samples

Problem

Review research is fragmented across platforms with different delivery models: dynamic Next.js data on Trustpilot, paginated tokens on Google Play, and limited page-rendered public review data on the App Store.

Solution

Built isolated scrapers for each source, then normalized review text, ratings, authors, dates, metadata, and source URLs into analysis-friendly JSON with deterministic deduplication and committed samples.

Key decisions

Kept one script per platform so source-specific fetching, pagination, and fallback behavior remain isolated and maintainable.
Used a Patchright browser context for Trustpilot before reading public Next.js JSON, reusing a realistic session without scraping visual markup.
Prioritized App Store page-rendered JSON and retained the legacy RSS path only as a fallback because the public RSS feed is often empty.
Preserved source URLs, deterministic review IDs, and fallback text/date keys so records remain auditable and deduplicated across runs.
Stored trimmed portfolio samples separately from full generated outputs and documented platform limitations rather than hiding or estimating missing coverage.

Tech stack

Python 3.11+
Patchright
google-play-scraper
Apple iTunes Lookup API
App Store public web JSON
JSON
PowerShell
Google-style docstrings

Key decisions

Kept one script per platform so source-specific fetching, pagination, and fallback behavior remain isolated and maintainable.

Used a Patchright browser context for Trustpilot before reading public Next.js JSON, reusing a realistic session without scraping visual markup.

Prioritized App Store page-rendered JSON and retained the legacy RSS path only as a fallback because the public RSS feed is often empty.

Preserved source URLs, deterministic review IDs, and fallback text/date keys so records remain auditable and deduplicated across runs.

Stored trimmed portfolio samples separately from full generated outputs and documented platform limitations rather than hiding or estimating missing coverage.