Facebook Group Finder with Fit Score

Type in a few keywords like boutique owners or keto recipes and get back a list of public Facebook groups about them. Each group comes with its member count, posts per day and per month, its description, and a 0–100 fit score that tells you how well it matches your niche.

It's built for community marketers, and it's also useful for agencies and coaches who need to find out where their audience already hangs out.

• No Facebook login, cookies or account needed
• Several keywords per run, and multi-word phrases work
• Groups found by more than one keyword show up only once
• The fit score is transparent and deterministic, and every part of it is explained below
• Only group-level metadata is collected, never anything about members (see Privacy)

How it works

1. Discovery. Facebook's own group search only works when you're logged in, so it isn't used. Instead, the Actor runs two search-engine queries for each keyword (site:facebook.com/groups "keyword" and "keyword" facebook group) and reads several result pages. It keeps every link that points to facebook.com/groups/....
2. Group details. For each group it opens the group's public About page through a residential proxy. Facebook shows that page to logged-out visitors. From it the Actor reads:
   • name, privacy, member count and post activity
   • description, topics, locations and creation date
3. Filter and score. It applies your privacy and minimum-members filters, drops groups that don't match any keyword, calculates the fit score and saves each group.

Input

Field	What it does	Default
keywords	Topics to search. Phrases are fine.	boutique owners, shopify store owners
maxResultsPerKeyword	Maximum number of groups to return per keyword	20
privacyFilter	PUBLIC returns public groups only. ANY also includes private groups, whose size, activity and description Facebook still shows publicly.	PUBLIC
minMembers	Skips groups smaller than this. Groups with an unknown size are also skipped when this is set.	empty
nicheDescription	Optional sentence about your audience or offer. It feeds 30% of the relevance score.	empty
requireKeywordMatch	Keeps a group only if its name, description or topics contain every distinctive word of the keyword it was found for (see Fit score). You aren't charged for skipped groups.	on
searchEngine	How groups are found (see below).	google_actor
maxSearchPagesPerQuery	Number of result pages read per query (2 queries per keyword). Leave it empty for automatic: 3 pages when maxResultsPerKeyword is 10 or less, 5 pages above that. More pages can find more groups but cost more (see Costs).	auto
proxyConfiguration	Proxy used for Facebook requests. Keep it on Residential.	Residential
healthCheck	Monitoring mode (see Health check)	off

Search engine options

• google_actor calls Apify's Google Search Scraper Actor. This is the recommended default. That run is started in your Apify account and billed separately (see Costs).
• google uses Google through Apify Proxy's Google SERP group. It doesn't start an extra Actor, but in testing it was slower and found fewer groups.
• duckduckgo is free but often blocked by DuckDuckGo's bot check.

If the chosen engine finds nothing for a keyword, the Actor tries DuckDuckGo once.

Example:

{
    "keywords": ["boutique owners", "shopify store owners"],
    "maxResultsPerKeyword": 20,
    "privacyFilter": "PUBLIC",
    "minMembers": 1000,
    "nicheDescription": "online clothing boutique owners looking for more customers"
}

Output

You get one dataset item per group. Any value Facebook doesn't show publicly is null; nothing is estimated or guessed. This example is built from the "Foodie" group's public About page as it was fetched on Oct 30, 2026, scored for the keywords ["foodie", "food lovers"]. The description is shortened here, and discoveredVia and scrapedAt are illustrative:

{
    "groupId": "330526660347434",
    "name": "Foodie",
    "url": "https://www.facebook.com/groups/330526660347434/",
    "privacy": "PUBLIC",
    "visibility": "VISIBLE",
    "memberCount": 269784,
    "memberCountIsApproximate": false,
    "memberCountText": "269,784 total members",
    "postsPerDay": 333.33,
    "postsPerMonth": 10000,
    "postsLastDay": 0,
    "activityText": "No posts today · 10.0K in the last month",
    "activityTextSource": "page_text",
    "activityStatus": "active",
    "description": "We Are not food geek, but We really love to eat...... This group is made out of love for food. ...",
    "location": null,
    "topics": null,
    "createdDate": "2012-04-17",
    "createdText": "Group created on April 17, 2012",
    "matchedKeywords": ["foodie"],
    "fitScore": 98,
    "fitScoreBreakdown": { "relevance": 40, "size": 22.6, "activity": 25, "accessibility": 10 },
    "keywordRelevance": [
        { "keyword": "foodie", "score": 1 },
        { "keyword": "food lovers", "score": 0.32 }
    ],
    "nicheCoverage": null,
    "discoveredVia": "google_actor",
    "scrapedAt": "2026-10-30T22:00:00.000Z"
}

Field notes:

• postsPerMonth is Facebook's own "posts in the last month" number. postsPerDay is postsPerMonth / 30, rounded to 2 decimals.
• activityText is Facebook's visible wording when the page includes it (activityTextSource: "page_text"). Otherwise it's spelled out from Facebook's own post counts ("structured_data"). If Facebook shows no activity at all, the counts are null and the text says "Activity not shown publicly by Facebook" ("not_shown").
• activityStatus is active when Facebook reports posts in the last month, dormant when Facebook itself reports 0 posts in the last month, and unknown when it shows no activity. A 0 is only ever Facebook's own figure, never a guess.
• memberCountIsApproximate is true only when Facebook showed a rounded figure such as "104.6K members" and no exact count.
• location lists the places the group names in its About section, or null when there are none.
• createdDate is the creation date Facebook shows on the About page.

You can export the results as CSV, Excel or JSON from the Groups view, or from All fields. A DIAGNOSTICS record in the key-value store explains every run, including how many pages were read, blocked or filtered.

A run never "succeeds" with zero results. If nothing is found, the run fails with a status message that says why: the search was blocked, no groups were found, Facebook showed a login wall, or everything was filtered out. You aren't charged for results on those runs.

Costs

A run has two separate costs:

1. This Actor: $0.008 per group saved to the dataset (the group-result event, $8 per 1,000 groups), plus Apify's standard Actor start event ($0.00005 per run at the default 1 GB memory). Filtered, failed or skipped groups aren't charged. Platform usage of this Actor's own run is included in the price.
2. Google Search Scraper (only with searchEngine: google_actor, the default): each run starts one run of apify/google-search-scraper in your own Apify account. It's billed to you separately, at that Actor's own pay-per-event price, and shows up as its own run in your Console.

The Google part depends on the number of result pages: pages = keywords × 2 queries × pages per query. With the automatic page setting, that's 6 pages per keyword when you ask for up to 10 groups per keyword, and 10 pages per keyword above that.

Approximate Google Search Scraper cost per keyword, from its price list as shown on Oct 30, 2026 ($0.0045 per page on the Free plan, $0.0025 on Bronze, plus $0.00005 per run). Check its Pricing tab for current prices.

Groups per keyword	Pages per keyword	Free plan	Bronze
up to 10 (3 pages per query)	6	about $0.027	about $0.015
more than 10 (5 pages per query)	10	about $0.045	about $0.025

Measured in our tests on Oct 30, 2026 (Bronze): 3 keywords at 5 groups each used 18 Google pages ($0.045); 1 keyword at 20 groups used 10 pages ($0.025). Search results change over time, so these are examples, not guarantees.

To keep costs predictable, set a maximum cost per run when you start the Actor. The Actor runs with 1 GB of memory by default (maximum 2 GB); it doesn't need more.

Fit score (0–100)

The fit score is the sum of four parts. It's deterministic: the same group data and input always give the same score.

Part	Points	Formula
Relevance	0–40	40 × R (see below)
Size	0–25	25 × min(1, log10(members + 1) / 6). 1,000,000 members earns the full 25.
Activity	0–25	25 × min(1, log10(postsPerMonth + 1) / log10(301)). 300 or more posts a month (about 10 a day) earns the full 25.
Accessibility	0–10	Public = 10, Private but visible = 4, otherwise 0

How R is calculated. Words are lower-cased, simple plurals are trimmed and common filler words are ignored ("group", "community", "the", and so on). Each keyword's words are then split into distinctive words and generic ones such as "recipes", "owners", "store", "business", "tips" or "beginners". In "keto recipes" the distinctive word is "keto"; in "shopify store owners" it's "shopify". If a keyword is made only of generic words, all of them count as distinctive.

• Each keyword gets a score from 0 to 1:
  • 0 unless every distinctive word appears somewhere in the group's name, description or topics. A shared generic word alone, like "recipes", doesn't count.
  • 1.0 if the whole keyword phrase appears in the group name
  • otherwise 0.75 if the phrase appears in the description or topics
  • otherwise 0.6 × (weighted share of keyword words in the name) + 0.4 × (weighted share in the description and topics), where distinctive words weigh 1 and generic words 0.25
• Combining keywords: K is the best score across your keywords.
• Without a niche description: R = K.
• With a niche description: R = 0.7 × K + 0.3 × min(1, 2 × N), where N is the share of your niche words found in the group's name, description and topics.

If the member count or the post count is null, that part scores 0; it's never estimated. The output includes fitScoreBreakdown and a per-keyword keywordRelevance, so you can see exactly where each score came from.

The score is a sorting aid, not a quality judgement. A small group with high relevance can be a better fit for you than a huge general one.

Privacy

This Actor only collects group-level public metadata, meaning the information Facebook shows about the group itself on its public About page to anyone who isn't logged in.

• It does not collect member lists, admins or moderators, post authors, commenters, profile links, or any other personal data.
• It does not log in, use cookies, or join groups.
• It does not read the contents of posts. Activity comes only from Facebook's aggregate counts.

You're responsible for how you use the results. Respect each group's rules and Facebook's terms, and don't spam groups.

Health check

Set "healthCheck": true to run three fixed keywords: boutique owners, shopify store owners and keto recipes, with up to 3 groups each and privacy set to ANY.

In this mode the Actor saves no dataset items, so there are no result charges. It writes a HEALTH_CHECK record (ok, emptyKeywords, sample rows and stats) and fails if any keyword returns zero groups. You can schedule it daily on Apify and alert on failed runs.

FAQ

Why not use Facebook's own group search? It needs a logged-in account, and logged-out requests to it return "not found". This Actor never uses an account.

Why do I need a residential proxy? Facebook shows a login wall to datacenter IPs and to any IP that views more than a few pages. In our tests, datacenter proxy requests hit the login wall on every group page we tried, while residential requests got through. With a residential proxy, every retry uses a new IP.

Why does a very busy group show exactly 10,000 posts a month? Facebook seems to cap or round very high activity figures (it shows "10.0K"). Treat 10,000 as "10,000 or more". This hasn't been fully verified.

Why does a big group show 0 posts a month? Because Facebook's own About page says so: it shows "No posts today" and "No posts in the last month", and its post counts are 0. These groups are flagged activityStatus: "dormant". Many are large but inactive, or have been renamed. If Facebook shows no activity at all, the counts are null instead.

Does it find every group on a topic? No. It finds the groups that search engines have indexed, so niche or new groups can be missing. To find more, add related keywords such as "boutique owners", "boutique wholesale" and "online boutique", or raise maxSearchPagesPerQuery.

Why are some groups not about my keyword? Search engines sometimes return loosely related groups. requireKeywordMatch (on by default) removes groups that don't contain your keyword's distinctive words ("keto" in "keto recipes"), and the relevance part of the score ranks the rest.

Are private groups included? Only when privacyFilter is ANY. Facebook still shows the size, activity and description of private groups publicly, but you'd need to join before you could post.

How am I charged? You're charged $0.008 per group saved to the dataset (the group-result event), plus a $0.00005 Actor start event per run. Groups that are filtered out, failed or skipped aren't charged, and the run stops at your maximum cost per run. With searchEngine: google_actor, the Google Search Scraper run is started in your account and billed separately. See Costs.

What if a run finds nothing? It fails with a clear message. See the DIAGNOSTICS record for details.