Scrapers

Business research

Deep 11-dimension business research. Given a domain or company name, produce a structured dossier covering identity, market, product, traction, capital, team, sentiment, technology, distribution, risk, and outlook. Output feeds sales intelligence and the discovery aggregator's brain ingest.

Updated today

Overview

At a glance

Worker: ultron-research
Pattern: Composite + LLM summarisation per layer
Input: One company target per run; many targets per task
Layers: 11
Latency: Roughly 60 to 180 seconds per target
Trigger: Cortex persona, leadgen enrichment, manual chat invocation

Most upstream scrapers return a single slice of a company. Maps gives places, similarweb gives traffic, research the actor gives a single composite dossier where every section is normalised. The shape is stable across targets so downstream consumers can rely on per-layer schemas.

The 11 layers

#	Layer	What it covers
1	Identity	Legal name, domain, founded year, HQ, legal form, LEI when available.
2	Market	Industry tags, sub-industry, geo footprint, TAM hints, category position.
3	Product	Primary product, key features, pricing model, tiers, target buyer.
4	Traction	Customers, ARR or revenue band when public, growth rate hints.
5	Capital	Funding stage, total raised, last round details, lead investors.
6	Team	Headcount band, leadership (CEO, founders), key recent hires.
7	Sentiment	Public reviews (G2 + Capterra summary), customer NPS hints when surfaced.
8	Technology	Tech stack hints, hiring signals, security posture clues.
9	Distribution	Marketing channels in use, social presence depth, content cadence.
10	Risk	Recent press incidents, regulatory exposure, named litigation if public.
11	Outlook	Synthesis: positioning, momentum, opening for outreach.

Input schema

actor inputjson

1{

2 "target": {

3 "domain": "linear.app", // preferred

4 "name": "Linear", // fallback when no domain

5 "countryHint": "US"

6 },

7 "layers": ["identity", "market", "product", "capital", "team", "outlook"], // optional; default all 11

8 "depth": "standard" // light | standard | deep (controls LLM token budget)

9}

Pipeline flow

pipelinetxt

1# Per target:

2# 1. Resolve target. If only name + country given, try to discover a

3# canonical domain via web search.

4# 2. Fan out per layer:

5# - identity, market, product -> web search + wellfound + pitchbook

6# - traction -> web search + customer-logo mining

7# - capital -> pitchbook + dealroom + crunchbase web

8# - team -> linkedin (auth) + wellfound

9# - sentiment -> g2-reviews + capterra-reviews

10# - technology -> similarweb + jobs (tech in postings)

11# - distribution -> similarweb + social profile scans

12# - risk -> web search (incidents, lawsuits)

13# 3. For each layer, the gathered raw rows feed a layer-specific

14# summariser prompt running on a lightweight model. Outputs a

15# structured per-layer object.

16# 4. The outlook layer is a synthesiser pass over the other 10.

17# 5. Push one composite kind="dossier" row to the dataset.

Output shape

dataset-row.jsonjson

1{

2 "kind": "dossier",

3 "source": "research",

4 "target": { "domain": "linear.app", "name": "Linear" },

5 "identity": {

6 "legalName": "Linear Orbit, Inc.",

7 "domain": "linear.app",

8 "founded": "2019",

9 "hq": "San Francisco, CA",

10 "legalForm": "Corporation"

11 },

12 "market": {

13 "industry": "B2B SaaS",

14 "subIndustry": "Project Management",

15 "geos": ["US", "EU"],

16 "categoryPos": "Challenger"

17 },

18 "product": {

19 "primary": "Issue tracking for software teams",

20 "pricingModel": "Per-seat",

21 "tiers": ["Free", "Standard", "Plus", "Enterprise"]

22 },

23 "traction": {

24 "customerHints": ["OpenAI", "Superhuman", "Ramp"],

25 "growthHint": "Reported 3x YoY headcount through 2024-2025"

26 },

27 "capital": {

28 "stage": "Series C",

29 "totalRaisedUsd": 130000000,

30 "lastRound": { "stage": "Series C", "year": 2024, "leadInvestor": "Accel" }

31 },

32 "team": { "headcountBand": "51-200", "leadership": [{ "title": "CEO", "name": "Karri Saarinen" }] },

  "sentiment":       { "g2Rating": 4.7, "capterraRating": 4.6, "summary": "Strong UX praise; reporting limitations called out." },

34 "technology": { "stackHints": ["TypeScript", "Postgres"], "hiringSignals": ["backend", "infra"] },

35 "distribution": { "channels": ["product-led", "community", "content"], "socialDepth": "Heavy on X and LinkedIn" },

36 "risk": { "incidents": [] },

  "outlook":         { "summary": "Established challenger with strong product loop and category momentum. Best opening: integrations or pricing-tier expansion.", "momentum": "high" },

38 "scrapedAt": "2026-05-19T12:00:00Z"

39}

Webhook contract

webhookjson

1{

2 "eventTypes": ["ACTOR.RUN.SUCCEEDED", "ACTOR.RUN.FAILED", "ACTOR.RUN.TIMED_OUT", "ACTOR.RUN.ABORTED"],

3 "requestUrl": "https://scraper-webhook.51ultron.com/apify",

4 "payload": { "actorName": "research", "source": "research", "datasetId": "<id>" }

5}

File map

actors/research

.actor

actor.json

INPUT_SCHEMA.json

src

__main__.pytarget resolver, layer fan-out, synthesiser

layers

identity.py

market.py

product.py

traction.py

capital.py

team.py

sentiment.py

technology.py

distribution.py

risk.py

outlook.py

Dockerfile

requirements.txt

README.md