Competitor Review Intelligence – G2, Trustpilot, Capterra & Bol.com

Turn competitor reviews into structured competitive intelligence. Which features drive 1-star ratings? Who are customers switching to? Which product gaps are your competitors failing to fix? AI-classified at $0.015/review instead of $300/month.

Features

• Multi-platform scraping — G2, Trustpilot, Capterra, and Bol.com in one run
• AI aspect classification — GPT-4o-mini classifies each review by feature: pricing, UX, support, integrations, performance, onboarding, reliability (multilingual — works on Bol.com's Dutch/French reviews with no special handling)
• Churn signal detection — flags reviews with explicit competitor-switching intent ("moved to", "switching to", "migrated to")
• Competitor name extraction — identifies which alternatives reviewers mention
• Star rating filters — focus on 1-2 star reviews for product gap research, 4-5 star for praise mapping

Inputs

Field	Default	Description
products	[]	Array of {name, g2Url?, trustpilotUrl?, capterraUrl?} — required unless bolComConfig is set
platforms	["g2","trustpilot"]	Platforms to scrape. sources is an alias (below), taking precedence if both set
sources	—	Alias for platforms; include "bol_com" here to enable Bol.com
bolComConfig	—	{ productUrls?: string[], searchKeyword?: string, maxReviewsPerProduct?: number } — at least one of productUrls/searchKeyword needed
maxReviewsPerProduct	100	Max reviews per product per platform
classifyAspects	true	Enable AI aspect classification
openaiApiKey	—	Required for AI classification
detectChurnSignals	true	Flag competitor-switch mentions
filterMinRating	1	Minimum star rating to process
filterMaxRating	5	Maximum star rating to process

Output fields

Each review record includes: productName, platform, rating, body, aspects[], primaryComplaint, primaryPraise, hasChurnSignal, churnReason, mentionedAlternatives, overallSentiment, sentimentScore, reviewerName, reviewerRole. Bol.com reviews additionally carry pros[], cons[], language ("nl"/"fr"), and priceAtReview.

Pricing (PPE)

• run-started — £0.05 per run
• review-scraped — £0.003 per raw review collected
• review-classified — £0.015 per AI-classified review
• churn-signal-detected — £0.025 per churn-intent review

Notes

G2 uses bot protection (Datadome). For best results, configure Apify residential proxies in your account settings. Trustpilot is scraped via the memo23/trustpilot-scraper-ppe sub-actor — live-verified during the Bol.com addendum build (this actor had never actually been run live before; two real pre-existing bugs were caught and fixed in the process: a missing ApifyClient token, and a reference to a Trustpilot sub-actor slug — nikita.rol/trustpilot-reviews-scraper — that no longer exists on the Apify Store).

Bol.com — known limitation (read before relying on this source)

Bol.com is protected by Akamai's TLS-fingerprint-level bot detection (JA3/JA4) — a layer below the browser that Crawlee's fingerprint spoofing (which successfully passes the Cloudflare-class JS challenges used by other sources in this account's actors) cannot touch. Every request currently receives an immediate 403 with no challenge page.

Two real proxy configurations were tried live on the deployed platform (this Apify plan disables local proxy testing entirely):

1. An explicit RESIDENTIAL proxy group — this account has 0 available RESIDENTIAL sessions; still 403 on every retry.
2. Automatic/default proxy selection (in case it routes through the account's enabled PROXY_UNBLOCKER feature) — still 403 on every retry.

What this means in practice: Bol.com scraping is built and wired in end-to-end (search, bulk product URLs, JSON-LD + Bazaarvoice review parsing, NL/FR language detection, price extraction), but currently returns 0 reviews on this account, because Bol.com's page HTML is never actually reachable. This is not a fake or partial feature — it's real, correct extraction logic (built from Bol.com's confirmed real schema.org JSON-LD and Bazaarvoice CSS conventions) that will work the moment page access is available, without further code changes — e.g. a RESIDENTIAL proxy allocation, the correct PROXY_UNBLOCKER group name (undocumented, not found), or Bol.com's official partner API (real, confirmed via their CSP header, but requires a registered partner/seller account).

Existing G2, Trustpilot, and Capterra scraping is unaffected by this — verified unchanged.