apify-google-maps-leads

Ask three questions as one block — don't drip them one by one.

1. Target audience — the business type(s) to search for. Free text. Examples: "dentists", "vegan restaurants", "boutique hotels", "dog groomers, pet stores" (comma-splittable → array).
2. Target geography — one location per run. City + country reads best ("Berlin, Germany", "Austin, TX"). If the user gives a country only, warn that Maps runs perform best on city-scoped searches.
3. Use reviews for lead scoring? (y/n) — if yes, we pull reviews (maxReviews) and filter by review volume + star rating post-run. If no, we skip reviews to cut cost.

Follow-ups only if the user asks for more control:

• maxCrawledPlacesPerSearch — default 50. Bigger runs = bigger cost.
• maximumLeadsEnrichmentRecords — default 3 per place (people to enrich per business). Never 0 — that disables leads enrichment, which is the point.
• leadsEnrichmentDepartments — default [] (any). Enum values are listed in references/actor-inputs.md.
• Minimum star rating pre-filter (placeMinimumStars) — cheaper than post-filtering when scoring is off.
• Review-scoring thresholds — default is ≥ 10 reviews AND ≥ 4.0 rating.

Cost warning threshold. Compute expected_leads = maxCrawledPlacesPerSearch × maximumLeadsEnrichmentRecords. If it exceeds 200, restate the number back to the user and confirm before running. Leads enrichment is the dominant cost line; a slip here is what surprises people.

Set every field below on every run. Full field reference in references/actor-inputs.md.

scrapeSocialMediaProfiles auto-enables scrapeContacts. Both are billed on top of the base scrape — see the Actor's pricing tab.

apify actors call "compass/crawler-google-places" \
  --input '<JSON_FROM_STEP_2>' \
  --user-agent apify-awesome-skills/apify-google-maps-leads \
  --json 2>/dev/null

Capture id (runId) and defaultDatasetId. Pull the dataset:

apify datasets get-items <DATASET_ID> --format json \
  --user-agent apify-awesome-skills/apify-google-maps-leads 2>/dev/null > places.json

Leads enrichment adds 30–90 s per place — expect long runs. If a run times out, the dataset already holds partial results; pull by datasetId.

Applied to the raw places.json in order. Full logic in references/scoring-and-backfill.md.

1. Spurious-match filter (always on). Drop leads whose leadsEnrichment[].companyWebsite hostname doesn't match the place.website hostname. Same failure mode as apify-verified-email-finder — a global-fallback lead attributed to unrelated places by substring. Count drops in run_metadata.json.
2. Review-based score (only if scoring is on).
   • Default keep-logic: place.reviewsCount >= 10 AND place.totalScore >= 4.0.
   • Emit a numeric Lead Score column: round(place.totalScore * log10(place.reviewsCount + 1), 2). Higher = better local reputation.
   • If scoring is off, Lead Score is blank.
3. Empty-lead surfacing. If a place has zero enriched leads, keep one row for the place with blank person fields — the user sees the business but knows nobody was found. Never silently drop.

Scalelist needs a person name (or LinkedIn URL) to look anything up. When a place has zero enriched leads — or leads with blank firstName / lastName — but has a working website, run apify/ai-web-scraper on that website to extract owner/decision-maker names.

When to run this step per place (all conditions must hold):

• place.website is non-empty (nothing to scrape otherwise)
• Either leadsEnrichment[] is empty, or every lead has a blank firstName
• The place survived Step 4's scoring filter (don't spend on places we're going to drop)

Payload — same input pattern as the Apify AI Web Scraper "list of writers" example, retargeted from blog authors to local-business decision-makers:

{
  "startUrls": [{"url": "<place.website>"}],
  "extractionMode": "agentic",
  "prompt": "Find the owner, founder, or key decision-makers of this business. For each person, include their full name and job title. Prioritize pages like /about, /team, /contact, or the site footer.",
  "maxPagesToVisit": 20,
  "maxCrawlDepth": 3
}

Defaults trimmed vs. the blog example (which uses maxPagesToVisit: 100 / maxCrawlDepth: 5) — small-business sites are typically shallow, and this is one call per website.

apify actors call "apify/ai-web-scraper" \
  --input '{"startUrls":[{"url":"<place.website>"}],"extractionMode":"agentic","prompt":"...","maxPagesToVisit":20,"maxCrawlDepth":3}' \
  --user-agent apify-awesome-skills/apify-google-maps-leads \
  --json 2>/dev/null

One place per call — the Actor's crawl fans out from startUrls, so batching multiple business sites in one startUrls array would mix results. Run one call per place website. Parallelize across places if you have many.

Merge results back onto the place:

• The Actor returns rows with url, data, markdown. data holds the extracted people — expect { "people": [{ "name": "...", "jobTitle": "..." }] } or an array of such objects (LLM output shape varies).
• For each extracted person, split name on the last space into firstName / lastName.
• Append a new lead into place.leadsEnrichment[] with firstName, lastName, jobTitle, companyWebsite = place.website, and mark Backfill Source = "ai-web-scraper".
• Cap at 3 new leads per place — the Actor sometimes returns lots of tangential names (past employees, testimonial subjects).

Skip conditions:

• Skip entirely if the user opts out of name discovery up front (offer this as a follow-up when running large batches — this is the priciest step per place).
• Skip if place.website returns a redirect to a social profile (Facebook page, Instagram) — the AI scraper handles JS sites but Meta login walls will burn budget for nothing.

For every lead surviving Step 5, check what's missing.

Phone backfill. Collect leads with a non-blank name and a blank phone. Group into batches of 100. Payload for scalelist/phone-finder:

{
  "leads": [
    {"first_name": "...", "last_name": "...", "company_domain": "...", "linkedin_profile_url": "..."}
  ]
}

linkedin_profile_url alone is sufficient; otherwise supply first_name + last_name + company_domain (preferred) or company_name.

apify actors call "scalelist/phone-finder" \
  --input '{"leads":[...]}' \
  --user-agent apify-awesome-skills/apify-google-maps-leads \
  --json 2>/dev/null

Email backfill. Collect leads with a non-blank name and a blank email. Payload for scalelist/email-finder:

{
  "leads": [
    {"first_name": "...", "last_name": "...", "company_domain": "...", "company_name": "..."}
  ]
}

first_name + last_name are required; company_domain beats company_name for match rate.

apify actors call "scalelist/email-finder" \
  --input '{"leads":[...]}' \
  --user-agent apify-awesome-skills/apify-google-maps-leads \
  --json 2>/dev/null

Both Actors are pay-per-event: you're only charged for successful matches. Merge the returned phones/emails back onto the original leads by lowercased first_name + last_name + company_domain.

Skip conditions. Don't call scalelist if the lead has no first+last name AND no LinkedIn URL — nothing to look up. Don't call it if the user says "skip backfill" up front (surface this as an offer if the initial maximumLeadsEnrichmentRecords was high, since the cost stacks).

Full column schema in references/output-format.md. Twenty columns including Lead Score, Instagram Followers, Facebook Followers, Backfill Source.

• Dedupe by lowercased email where present; otherwise by lowercased first_name + last_name + place_id.
• Sort by Lead Score descending when scoring is on; otherwise by Business alphabetically.
• Deliverable header: state whether scoring was on, the review thresholds, and the spurious-match drop count. Offer to re-render without scoring if the kept-row count is low.
• Write a run_metadata.json sidecar next to the CSV with runId, datasetId, and stats (placesScraped, rawLeads, spuriousDropped, phonesBackfilled, emailsBackfilled, keptRows).