Comparison research

Product name in. Listing URLs out. Then a structured extract of ratings, reviewer themes, alternatives, and category peers. Do not mint a C_xxxxx profile, bind competitor_ids, or quote a rating on a call from this pass.

This page is the directory-site companion to Research. Canonical competitor profiles still need requester-supplied G2 and Capterra review links. See Competitors.

Tools

Same default stack as Research. These sites are JS-heavy and often return 403 to a naked GET. Render, then extract.

Tool	Agent surface	Use for
Cloudflare Browser Run	`/markdown`, `/json`, `/links`	Render the listing. Structured extract. Pagination links.
Firecrawl	`/v2/search` with `includeDomains`, `/v2/scrape` JSON, `/v2/map`	Resolve the real listing URL. Scrape reviews. Map `/reviews` and `/alternatives`.
Exa	`/search` `includeDomains`, `/contents`	Backup discovery when Firecrawl search is thin.

POST https://api.cloudflare.com/client/v4/accounts/<accountId>/browser-rendering/{markdown|json|links}
POST https://api.firecrawl.dev/v2/{search|scrape|map}

Default loop: Firecrawl /search to resolve the listing. Browser Run or Firecrawl /scrape to read it. Stop on a challenge, login, or paywall. Do not retry with a different fingerprint, cookie, or solver.

Context

Use this page when the job is:

Find this product on G2-class sites.
Pull displayed ratings, review counts, and reviewer themes.
Pull the alternatives / "vs" graph the directory already computed.
Pull category peers (who else sits on the same grid).

Do not use it to invent a competitor, to average two sites into one score, or to treat a reviewer sentence as product behavior.

Axioms

Public pages only. No account, no "unlock reviews," no bot-check bypass. If the tool returns a challenge or gate, record blocked and stop that site.
Never guess a Capterra numeric id, a G2 slug, or a rating. Resolve the URL from search. Copy metrics as displayed, with retrieved_at.
Say reviewers describe / one reviewer says. Keep site scores separate. Do not blend G2 and Capterra.
Cap volume. Three review pages or ~50 reviews per site unless a person asks for more. Prefer newest first.
Gartner Digital Markets (Capterra, GetApp, Software Advice) and Slashdot Media (SourceForge, Slashdot Software) syndicate. Do not count the same review three times.
A listing URL is a candidate for the competitor workflow. It does not replace requester-supplied G2 and Capterra links.

1. The 20 sites

Priority order for a sweep, not a traffic ranking. Always run 1–2. Run 3–9 when you need themes or an alternatives graph. Run 10–18 when 1–9 are thin. Run 19–20 when the vendor or buyer is EU/DACH.

#	Site	Apex	Job	Skip when
1	G2	`g2.com`	Reviews, grids, alternatives, vs pages. Canonical for profiles.	—
2	Capterra	`capterra.com`	Reviews, pricing hints, alternatives. Canonical for profiles.	—
3	TrustRadius	`trustradius.com`	Longer verified reviews, scorecards, comparisons.	—
4	PeerSpot	`peerspot.com`	Enterprise IT comparisons, "vs," peer Q&A.	Consumer / SMB-only tools
5	Gartner Peer Insights	`gartner.com/reviews`	Enterprise peer ratings by market.	Full text often gated
6	SoftwareReviews (Info-Tech)	`softwarereviews.com`	Data Quadrant, capability scores.	Report PDF gated
7	SelectHub	`selecthub.com`	Requirements matrices, category shortlists.	—
8	FeaturedCustomers	`featuredcustomers.com`	Vendor-verified customer reviews and case studies.	No vendor page
9	AlternativeTo	`alternativeto.net`	Crowd alternatives graph, not review depth.	—
10	Software Advice	`softwareadvice.com`	Advice guides + reviews. Gartner Digital Markets.	Capterra already scraped
11	GetApp	`getapp.com`	Directory + reviews. Gartner Digital Markets.	Capterra already scraped
12	SourceForge	`sourceforge.net/software`	Directory, reviews, category lists. Slashdot Media.	—
13	Slashdot Software	`slashdot.org/software`	Same family as SourceForge. Category compare.	SourceForge already scraped for the same listing
14	Clutch	`clutch.co`	B2B services and SIs, not the ERP SKU. Use for implementers.	Product-only research
15	GoodFirms	`goodfirms.co`	Software + agency directory, reviews.	—
16	FinancesOnline	`financesonline.com`	Editorial score + user reviews.	Treat editorial as editorial
17	Crozdesk	`crozdesk.com`	Category score, alternatives.	—
18	SaaSworthy	`saasworthy.com`	Feature comparison tables, alternatives.	—
19	OMR Reviews	`omr.com`	DACH/EU user reviews.	Non-EU vendor, no listing
20	Appvizer	`appvizer.com`	FR/EU directory, comparisons, guides.	Non-EU vendor, no listing

Worked seed we already cite: Global Shop G2 reviews (read 2026-08-02 on the canonical profile). Resolve every other listing from search. Do not guess a Capterra /p/{id}/.

2. Resolve the listing URL

Never construct Capterra ids. Never assume the G2 slug is the company legal name. Search, then pick.

POST /v2/search
{
  "query": "\"{product name}\" reviews",
  "includeDomains": ["{apex}"],
  "limit": 5
}

Keep the hit whose path matches the site's listing shape below. Drop blog posts, ads, /compare/ pages (unless the job is vs), and /categories/ pages (unless the job is peers).

If zero hits: not_found. Do not pick a similarly named product.

If several products share a vendor, keep the SKU the buyer would name (ERP vs HCM vs a module). Record the URL you rejected.

3. URL cookbook

Templates are for after search returns a live slug or id. {slug} is whatever the directory used, not our ID stem.

g2.com
  product:        /products/{slug}
  reviews:        /products/{slug}/reviews
  reviews_page:   /products/{slug}/reviews?page={n}
  alternatives:   /products/{slug}/competitors/alternatives
  compare:        /compare/{slug-a}-vs-{slug-b}
  category:       /categories/{category}

capterra.com
  product:        /p/{id}/{slug}/
  reviews:        /p/{id}/{slug}/reviews/
  alternatives:   /p/{id}/{slug}/alternatives/
  category:       /{category}-software/
  search:         /search/?query={q}

trustradius.com
  product:        /products/{slug}
  reviews:        /products/{slug}/reviews
  competitors:    /products/{slug}/competitors
  compare:        /compare/{slug-a}-vs-{slug-b}

peerspot.com
  product:        /products/{slug}-reviews
  compare:        /products/comparisons/{slug-a}-vs-{slug-b}
  category:       /categories/{category}

gartner.com
  market:         /reviews/market/{market-slug}
  vendor:         /reviews/vendor/{vendor-slug}
  product:        /reviews/market/{market-slug}/vendor/{vendor}/product/{product}

softwarereviews.com
  product:        /products/{slug-or-id}
  category:       /categories/{category}

selecthub.com
  product:        /{category}/{slug}/
  category:       /{category}/

featuredcustomers.com
  vendor:         /vendor/{slug}
  reviews:        /vendor/{slug}/reviews

alternativeto.net
  software:       /software/{slug}/
  about:          /software/{slug}/about/

softwareadvice.com
  profile:        /{category}/{slug}/
  category:       /{category}/

getapp.com
  product:        /{segment}-software/{slug}/   (path varies; trust search)

sourceforge.net
  product:        /software/{slug}/
  category:       /software/{category}/

slashdot.org
  product:        /software/{slug}/
  category:       /software/{category}/

clutch.co
  profile:        /profile/{slug}
  category:       /{category}

goodfirms.co
  software:       /software/{slug}

financesonline.com
  product:        /{slug}/

crozdesk.com
  product:        /{category}/{slug}

saasworthy.com
  product:        /product/{id}/{slug}

omr.com
  product_en:     /en/reviews/product/{slug}
  product_de:     /de/reviews/{slug}

appvizer.com
  product:        /{lang}/{category}/{slug}

After you have the product URL, Firecrawl map beats guessing extra paths:

POST /v2/map
{ "url": "{product_url}", "search": "reviews alternatives compare competitors" }

4. Extract

Render the page (Browser Run /json or Firecrawl scrape with formats: [{ type: "json", schema: ... }]). Copy visible text. If a field is not on the page, null. Do not fill from memory.

Listing card

One per site.

{
  "site": "g2",
  "url": "",
  "retrieved_at": "YYYY-MM-DD",
  "status": "ok | blocked | not_found | gated",
  "product_name": "",
  "vendor_name": "",
  "slug": "",
  "rating": null,
  "rating_scale": 5,
  "review_count": null,
  "categories": [],
  "description_on_page": "",
  "pricing_hint": null,
  "top_pros": [],
  "top_cons": []
}

rating and review_count must be the figures shown next to each other on that page. A star graphic without a count is review_count: null.

Reviews

Array, capped. Summarize body. Do not paste full review prose into a canonical profile later.

{
  "site": "g2",
  "url": "",
  "retrieved_at": "YYYY-MM-DD",
  "title": "",
  "score": null,
  "date": "",
  "reviewer_role": null,
  "industry": null,
  "company_size": null,
  "deployment": null,
  "pros": [],
  "cons": [],
  "themes": [],
  "body_summary": "",
  "verbatim_ok": false
}

themes are short labels you assign (scheduling, training_burden, reporting). body_summary is one or two sentences. Set verbatim_ok: true only when you kept a short attributed quote.

Pagination: Cloudflare /links or the rel=next URL on the page. Stop at 3 pages, ~50 reviews, or no next link.

Alternatives and vs

{
  "source_url": "",
  "seed_product": "",
  "others": [
    { "name": "", "slug": null, "url": "", "why_listed": "" }
  ]
}

why_listed is whatever the directory showed (category overlap, "also compared," switchers). It is not our judgment that they compete.

Category peers

From a category grid / shortlist / quadrant page:

{
  "site": "g2",
  "category_url": "",
  "category_name": "",
  "peers": [
    { "name": "", "listing_url": "", "rating": null, "review_count": null }
  ]
}

5. Jobs

Run one job per pass.

A. Listing sweep (default)

For sites in scope (always 1–2, then 3–9 as needed):

Resolve URL (§2).
Extract listing card.
If status is ok and the job needs themes, extract reviews with the cap.
Write one row per site into the table below.

site | status | listing_url | rating | review_count | retrieved_at | notes

B. Alternatives graph

From G2 /competitors/alternatives, Capterra /alternatives/, TrustRadius /competitors, AlternativeTo /software/{slug}/, plus any /compare/ links /links returns. Union by normalized name. Record every source URL. Do not drop a name because it is not in our catalog. Do not add it to the catalog from this job.

C. Category peers

Resolve the category slug from the listing card's categories[]. Scrape the category URL. Keep the visible leaderboard / grid, not ads.

D. Head-to-head vs

Only when you already have two slugs. Hit /compare/{a}-vs-{b} on G2 or TrustRadius, or PeerSpot /products/comparisons/. Extract the displayed score table. Still keep per-site figures. Do not publish "we beat them on G2" as a claim.

6. Agent runbook

product_name = <buyer-facing name>
sites = [g2, capterra] + optional
for site in sites:
  hits = firecrawl.search(query='"'+product_name+'" reviews', includeDomains=[apex])
  listing_url = first path that matches the cookbook
  if none: row.status = not_found; continue
  page = browser_run.json(listing_url, schema=listing_card)
  if challenge or login: row.status = blocked; continue
  if job includes reviews: scrape reviews pages 1..3
  sleep / rate-limit
write table
stop

If JSON names a product that does not appear in the markdown, discard the extract and re-read. Hallucinated ratings are worse than not_found.

Done when: every in-scope site is ok, not_found, gated, or blocked, each ok row has url + retrieved_at, and G2/Capterra URLs (if ok) are ready to hand to a person for the competitor workflow.

What this is not

Permission to write docs/sales/competitors/C_<nnnnn>_*.md or to supply guessed review links to that workflow.
Proof that a prospect uses the product. Directories list vendors, not our accounts. Customer URLs are Research and Domain scanning.
A single blended score. Two sites disagree → keep both.
License to scrape behind a login or to ignore a bot check.

Research — vendor sites, customer URLs, similar pages
GEO — if the engine cites G2, this scrape is the follow-on
AEO — developer comparison queries in coding harnesses
ERP signal research — 50–500 ERP need / want / window sources
Competitors
GTM guide structure — 1.3 market landscape and 2.4 battlecards request sourced content only

Change log

2026-08-31 — First page: 20 B2B comparison sites, listing resolver, URL cookbook, extract schemas, capped review scrape. Cloudflare Browser Run + Firecrawl default.

Tools​

Context​

Axioms​

1. The 20 sites​

2. Resolve the listing URL​

3. URL cookbook​

4. Extract​

Listing card​

Reviews​

Alternatives and vs​

Category peers​

5. Jobs​

A. Listing sweep (default)​

B. Alternatives graph​

C. Category peers​

D. Head-to-head vs​

6. Agent runbook​

What this is not​

Related knowledge​

Change log​