# Propensify data sources

Public-altitude guides for evidence Propensify can observe on a prospect company.
This document is generated at website build from the same content as https://propensify.io/data-sources.
It is not the Signal Catalogue. It does not list internal provider ids or scrape APIs.

Canonical HTML hub: https://propensify.io/data-sources
Compact bundle (LLM/planner): https://propensify.io/data-sources.compact.md
Full bundle: https://propensify.io/data-sources.md

## Contents

- [LinkedIn Ads](#linkedin-ads) — /data-sources/linkedin-ads
- [Google Ads](#google-ads) — /data-sources/google-ads
- [Meta Ads](#meta-ads) — /data-sources/meta-ads
- [Website signals](#website-signals) — /data-sources/website-signals
- [Website Ranking](#website-ranking) — /data-sources/website-ranking
- [Search agents](#search-agents) — /data-sources/search-agents
- [Companies House](#companies-house) — /data-sources/companies-house
- [Crunchbase](#crunchbase) — /data-sources/crunchbase
- [Firmographic & technographic](#firmographic-technographic) — /data-sources/firmographic-technographic
- [Job openings](#job-openings) — /data-sources/job-openings
- [Company workforce](#company-workforce) — /data-sources/company-workforce
- [Employee voice](#employee-voice) — /data-sources/employee-voice
- [Amazon Marketplace](#amazon-marketplace) — /data-sources/amazon-marketplace
- [Customer reviews](#customer-reviews) — /data-sources/customer-reviews
- [Company socials](#company-socials) — /data-sources/company-socials

---

# LinkedIn Ads

HTML: https://propensify.io/data-sources/linkedin-ads
Markdown: https://propensify.io/data-sources/linkedin-ads.md

## Three questions we ask of the ads

1. **Who are they aiming at?**
2. **What are they promising?**
3. **Where do they send you?**

No matching creatives in this pass is unknown — not that they never advertise.

## Presence is a gate. Destination is the angle.

### Presence & activity

### Destination & creative

## How ads evidence stacks

1. **Activity** — Live → historical
2. **Footprint** — Catalogue / impressions
3. **Destination** — Where ads send people
4. **Mix** — Creative & geo

## Activity

Start with activity. A large catalogue that isn’t active is a different conversation than ads running this week.

Live → historical

## Campaign activity

_3 intermediate states omitted for compact budget._

- **Active now**
- **Thin dating**

## Launch cadence

- **Accelerating**
- **Stable**
- **Decelerating**

## Footprint

## Catalogue Bands

- **Large**
- **Moderate**
- **Small**
- **Empty**

## Impression Bands

- **Very high**
- **High**
- **Mid**
- **Low**

## Where ads send people

Several destination types can fire on the same account.

## Destinations

- **Demo or trial**
- **Dedicated landing page**
- **Content / PDF**
- **Short-link heavy**

## Creative & CTAs

## Creative

- **Image-led**
- **Document-heavy**
- **Learn more CTAs**
- **CTA sparse**

## Geo reach

## Geo

- **Concentrated**
- **Multi-market**
- **Unknown**

## When signals stack

## Composites

- **Scaled and active**
- **Paid traffic into a landing path**
- **Content-led demand gen**
- **Historical only**

## Context we also surface

## Items

- **Destination URL profile**
- **Themes & keywords**
- **CTA inventory**
- **Audit sample**

## Boundaries & method

- Public Ad Library sample — not Ads Manager spend or exact impression totals.
- Impression fields are ranges; we use them as relative bands.
- We paginate the library (default several pages) — still incomplete when totals exceed what we fetch.
- Short links stay unclear until resolved — we don’t invent the landing page.
- No Ad Library ads ≠ “never advertises.” Pair with Website Ranking or hiring when the ICP needs traffic floor or capacity.

---

# Google Ads

HTML: https://propensify.io/data-sources/google-ads
Markdown: https://propensify.io/data-sources/google-ads.md

## Three questions we ask of the ads

1. **Which queries and messages?**
2. **What offer is the click for?**
3. **Which URL is doing the work?**

No matching creatives in this pass is unknown — not that they never advertise.

## Presence is a gate. Path is the angle.

### Presence & activity

### Path & surfaces

## How ads evidence stacks

1. **Activity** — Live → historical
2. **Footprint** — Catalogue / impressions
3. **Destination** — Where ads send people
4. **Mix** — Search · YouTube · Shopping

## Activity

Start with activity. A large catalogue that isn’t recently shown is a different conversation than ads live this week. Impressions may be missing even when activity is clear.

Live → historical

## Campaign activity

_3 intermediate states omitted for compact budget._

- **Active now**
- **Insufficient dates**

## Launch cadence & texture

_3 intermediate states omitted for compact budget._

- **Accelerating**
- **Burst / flight**

## Footprint

## Catalogue Bands

- **Large**
- **Moderate**
- **Small**
- **Empty**

## Impression Bands

- **Very high**
- **High**
- **Mid**
- **Low**

## Where on Google

Several surfaces can fire on the same account.

## Surfaces

- **Search-led**
- **YouTube present**
- **Shopping present**
- **Multi-surface**

## Where ads send people — and what they’re offering

Several destination and offer types can appear together.

## Destinations

- **Demo or trial**
- **Dedicated landing page**
- **Content / gated**
- **Pricing**

## Offers

- **Demo or trial**
- **Pricing**
- **Compliance**
- **Competitive**

## When signals stack

## Composites

- **Scaled and active**
- **Paid traffic into a landing path**
- **Search-led GTM**
- **Compliance-led demand**

## Boundaries & method

- Public Transparency sample — not Google Ads Manager spend, clicks, or exact impressions.
- Impression fields are ranges and often missing; Unavailable ≠ low spend.
- Catalogue may use a public estimate larger than fetched; pagination can be incomplete.
- Short links stay unclear until resolved — we don’t invent the landing page.
- No public Transparency ads ≠ “never advertises on Google” (non-public / login-gated ads exist).
- Domain seed quality matters; multi-advertiser IDs under one domain can mix attribution. Pair with Website Ranking or LinkedIn Ads when the ICP needs traffic floor or LinkedIn-specific GTM.

---

# Meta Ads

HTML: https://propensify.io/data-sources/meta-ads
Markdown: https://propensify.io/data-sources/meta-ads.md

## Three questions we ask of the library

1. **What claim is the creative making?**
2. **What motion is the CTA?**
3. **Is the program alive or collapsing?**

No ads in this pass is unknown — not that they do not advertise.

## Presence is a gate. Path is the angle.

### Presence & activity

### Path & surfaces

## How ads evidence stacks

1. **Activity** — Live → historical
2. **Footprint** — Catalogue / impressions
3. **Destination** — Where ads send people
4. **Mix** — Surfaces, creative & geo

## Activity

Start with activity. A large catalogue that isn’t active is a different conversation than ads running this week. Default searches are often active-only — historical inactive creatives may not be in the sample.

Live → historical

## Campaign activity

_3 intermediate states omitted for compact budget._

- **Active now**
- **Insufficient dates**

## Launch cadence & texture

_3 intermediate states omitted for compact budget._

- **Accelerating**
- **Burst / flight**

## Footprint

## Catalogue Bands

- **Large**
- **Moderate**
- **Small**
- **Empty**

## Impression Bands

- **Very high**
- **High**
- **Mid**
- **Low**

## Where on Meta

Several surfaces can fire on the same account.

## Surfaces

- **Instagram-led**
- **Instagram present**
- **Multi-surface**
- **Cross-surface always-on**

## Where ads send people — and what they’re offering

Several destination and offer types can appear together.

## Destinations

- **Demo or trial**
- **Dedicated landing page**
- **Content / gated**
- **Ecommerce**

## Offers

- **Content lead**
- **Demo or trial**
- **Webinar / event**
- **Ecommerce / product**

## Creative & CTAs

## Creative

- **Video-heavy**
- **Image-led**
- **DCO-heavy**
- **Hard conversion CTAs**

## Geo reach

## Geo

- **Concentrated**
- **Multi-market**
- **Unknown**

## When signals stack

## Composites

- **Scaled and active**
- **Paid traffic into a landing path**
- **Content-led demand gen**
- **Always-on content**

## Boundaries & method

- Public Meta Ad Library sample — not Ads Manager spend, ROAS, or exact impressions.
- Impression fields are often missing on commercial ads; Unavailable ≠ low spend.
- Default searches are frequently active-only — inactive historical creatives may not be analyzed.
- Pagination can be incomplete; we don’t claim the full catalogue was fetched.
- Wrong Facebook page seed is possible — page mismatch risk means treat ads as provisional.
- Short links stay unclear until resolved. Pair with Website Ranking or LinkedIn/Google Ads when the ICP needs traffic floor or other-channel GTM.

---

# Website signals

HTML: https://propensify.io/data-sources/website-signals
Markdown: https://propensify.io/data-sources/website-signals.md

## Three questions we ask of their own site

1. **Who do they say they are?**
2. **How do they sell, and what do they sell?**
3. **What do they admit in the margins?**

No matching sentence in this pass is unknown — not a no.

## Lanes

- **Publish**
- **Load**
- **Look**

## What they say. What the site loads. How it looks.

### What they say

### What the site loads

### How it looks

## How website evidence stacks

1. **Pages** — Families found on their site
2. **Offer cues** — Pricing · customers · team
3. **What loads** — Motion · tags · widgets
4. **Look** — Brand / layout when captured

## Page families

## Family Chips

- **Homepage**
- **About / story**
- **Team**
- **Products / services**

## Pricing & commercial cues

Public pricing paths also inform sales motion when we can read the site’s front door — see What the site loads. Site prices are marketing claims — not signed contract truth.

## Pricing Cards

- **Pricing page present**
- **No public pricing**

## Proof & team

## Proof Cards

- **Customers / proof**
- **Team / leadership**

## What the site loads

A partial read or a quiet checklist is not proof they use none of these products. Parked or unreadable pages are coverage — not a dead company, and not “no stack.” A tag manager is not proof of which tags fire inside it.

## Load Footprint

- **Domain doesn’t resolve**
- **Parked / for-sale**
- **Can’t read the page**
- **Checklist quiet**

## Sales Motion

Public conversion path

The site may also mark itself in structured markup (organization, software, local business) — a markup claim, not a legal classification.

Sales motion on the site

How they ask people to buy on the public site — self-serve, published pricing, demo, booking, or a contact form. Not the real enterprise procurement motion.

## Chips

- **Self-serve / try**
- **Public pricing path**
- **Demo / talk to sales**
- **Book a time**

## Stack families

Industry widgets appear when vertical schedulers or POS embeds show up — an embed is not proof of a paying tenant.

## Stack Families

- **Booking**
- **Payments**
- **Marketing automation**
- **Email / SMS**

## Also on the public site

## Stack Texture

- **Product analytics & experiments**
- **Publisher ads**
- **Cookie consent**
- **On-site reviews**

## More load texture

## Stack Soft Chips

- **Affiliate**
- **Call tracking**
- **Attribution**
- **Site search**

## Trust & compliance cues

## Trust Cards

- **Trust / security page**
- **Trust badge**
- **UK statutory page**

## App & surface hints

## App Hints

- **App link files**
- **Product host on certificate**

## Site inventory texture

## Inventory Soft

- **Sitemap size cue**
- **Sitemap freshness cue**

## Visual signals

Evidence-first visual notes. We do not invent psychographics or intent theatre from a screenshot.

## Visual Cards

- **Brand & layout cues**
- **Nested, not a separate guide**

## Boundaries & method

- Site claims ≠ Companies House registry or enrichment truth.
- Team pages ≠ Company workforce headcount series.
- Careers on site ≠ Job openings dataset (→ Job openings).
- Public pricing ≠ signed ACV or contract terms.
- Absence of a pricing page is observational — not proof of no commercial model.
- Customer logos are self-selected marketing proof.
- Custom themes are hypothesis-driven — not an unbounded whole-site index.
- What they publish and what the site loads are complementary — copy is not a substitute for loaded tags.
- Checklist quiet ≠ they use none of these products.
- Unreadable, parked, or unresolved ≠ the company is dead, and ≠ no stack.
- A tag or widget host ≠ spend, seats, or revenue.
- Trust badge or trust page ≠ SOC 2 or any attestation.
- UK statutory page ≠ filing contents.
- App-link files and certificate names ≠ installs; certificate hosts may be stale.
- Sitemap size ≠ traffic (→ Website Ranking). Sitemap dates are often stale or fake.
- Loaded-stack cues ≠ Firmographic technology name-lists.
- Ads pixels ≠ Google Ads, Meta Ads, or LinkedIn Ads libraries.
- Publisher ads (they sell ads on the page) ≠ Google Ads spend.
- On-site review widgets ≠ Customer reviews dataset.
- Tag manager present ≠ proof of which tags fire inside it.
- CMS or edge host ≠ exclusive hosting or WAF spend.
- An industry embed ≠ a paying tenant on a directory page.
- We do not list every library on the page.
- Visual cues are nested evidence — not a full brand audit.

---

# Website Ranking

HTML: https://propensify.io/data-sources/website-ranking
Markdown: https://propensify.io/data-sources/website-ranking.md

The floor number lives here. On-the-internet-at-scale corroboration is Ads presence and Company socials; hidden pages are Search agents.

- **Also read:** [Company socials](https://propensify.io/data-sources/company-socials) — public presence at scale, not the rank number
- **Also read:** [Website signals](https://propensify.io/data-sources/website-signals) — pages they publish, not the floor number

## Size is a floor. Movement is prioritization.

### Audience scale

### Ranking movement

## How ranking evidence stacks

1. **Scale** — Audience band
2. **Trajectory** — Improving → declining
3. **Composites** — Scale + movement together

## Audience scale

## Volume Bands

- **Very high**
- **High**
- **Mid**
- **Low**

## Ranking trajectories

Start with the shape. Scale tells you the floor; trajectory tells you whether to prioritize now.

Improving → declining

## Public popularity movement

_5 intermediate states omitted for compact budget._

- **Strong improvement**
- **Insufficient history**

## Scale + movement together

## Composites

- **Scaled and growing**
- **Scaled but declining**

## Boundaries & method

- Global web rank is a relative popularity proxy derived from public series — not first-party analytics or guaranteed visit counts.
- Lower global rank means a larger estimated audience; glyphs show popularity direction, not “rank number went up = good.”
- Research windows follow available history (commonly ~30 days on free coverage); longer history when the source allows.
- Unknown / empty series is a coverage miss — not proof of low traffic.
- Traffic is power-law distributed: compare bands, not tiny gaps between neighboring ranks.
- Website Ranking does not replace ads, hiring, or firmographic sources — it is often paired with them downstream.
- We prefer “insufficient data” over inventing monthly visit numbers.

---

# Search agents

HTML: https://propensify.io/data-sources/search-agents
Markdown: https://propensify.io/data-sources/search-agents.md

## Three questions we ask of the open web

1. **What did their nav hide?**
2. **Who talks about them off-domain?**
3. **What just happened in public?**

No result means we did not find it — not that it is false.

## Finding is cheap. Verifying is the filter.

### Find

### Verify

## How search evidence stacks

1. **Need** — What we’re looking for
2. **Where we look** — Hidden · off-domain · news
3. **Read** — Snippet vs full page

## Where we look

## Where Chips

- **Hidden pages**
- **Off-domain**
- **News**

## Open-web presence

A thin snippet can answer a simple need. It is not proof of everything on the full page.

## Open Web Cards

- **Owned pages found**
- **Snippet answers**
- **Page worth reading**
- **Nothing found**

## News & press

Empty news feed means nothing turned up in this sample — not “never in the news forever.”

## News Cards

- **News hits**
- **Topic-filtered news**
- **No articles in sample**

## Honesty

## Honesty Cards

- **Wrong entity risk**
- **Wrong topic risk**

## Boundaries & method

- Results are a sample — not a full web index or complete news archive.
- Snippet answers are thin; they are not the full page.
- Owned-page preference discards many third-party mentions from open-web evidence by design.
- Nothing found ≠ the topic never exists online.
- Wrong-entity and wrong-topic mismatches can still occur — honesty over false certainty.
- Search agents ≠ Website Ranking traffic floor.
- Search agents ≠ Google Ads inventory.
- Search agents ≠ Website signals page-family captures (pair those guides).
- No query-syntax or vendor infrastructure details on this page.

---

# Companies House

HTML: https://propensify.io/data-sources/companies-house
Markdown: https://propensify.io/data-sources/companies-house.md

## Two questions we ask of the filing

1. **Do they legally exist here?**
2. **Is there distress language in the filing?**

No filing in this pass is unknown — not that the company is empty.

## Identity is a gate. Depth is prioritization.

### Identity

### Depth

## How Companies House evidence stacks

1. **Identity** — Verified UK company
2. **Filings** — Disclosure gate
3. **Statutory numbers** — Parsed lines when present

## Entity & registry status

## Identity States

- **Verified**
- **Not found**
- **Ambiguous**

## Status States

- **Active**
- **Dissolved**
- **Insolvency**
- **Other / unclear**

## Overdue Card

Accounts overdue

Next accounts marked overdue on the profile — filing clock, not insolvency.

## Filings & disclosure

Filing found ≠ turnover known. Disclosure class decides whether a document pull is worth it.

## Filing Footprint

- **Accounts filing found**
- **None found**

## Disclosure class

- **Rich disclosure**
- **Low disclosure**
- **Unknown**

## Filing Flags

- **Paper-filed**
- **Amended**

## Statutory numbers

Other statutory lines attach as magnitudes when present — we don’t invent size bands from them.

## Parse Outcomes

- **Financials observed**
- **Low-disclosure skipped**
- **No filing**
- **Document unavailable**

## Headline Lines

- **Turnover disclosed**
- **Employees disclosed**
- **Net assets disclosed**

## Also Available Chips

- Cash
- Gross profit
- Profit before tax
- Inventories
- Debtors
- Creditors (≤1yr)
- Tangible fixed assets
- Total fixed assets
- Total assets

## Boundaries & method

- UK Companies House only — absence is not “no legal entity worldwide.”
- Registry status is not commercial activity or buying intent.
- Accounts overdue is profile filing metadata — not an insolvency verdict.
- Filing metadata is not parsed financials; magnitudes require a successful document parse.
- Low-disclosure classes often omit P&L/turnover; we may skip the document pull on purpose.
- Paper-filed / PDF text paths can yield “document obtained, numbers unavailable” — we fail closed.
- Statutory averages and GBP lines are observations — not credit scores or “can they afford us.”
- Officers, parent-group escalation, and keyword narrative stay in deeper research — this guide covers the public-register ladder.

---

# Crunchbase

HTML: https://propensify.io/data-sources/crunchbase
Markdown: https://propensify.io/data-sources/crunchbase.md

## Two questions we ask of the profile

1. **What capitalization story is on the profile?**
2. **What events are attached?**

No org page in this pass is unknown — not that they are unfunded.

## Snapshot is context. Momentum is prioritization.

### Snapshot

### Momentum

## How Crunchbase evidence stacks

1. **Profile** — Org snapshot
2. **Funding** — Rounds depth
3. **Growth** — Trends & milestones
4. **Tech & traffic** — Visits / stack

## Company profile

## Profile Footprint

- **Org observed**
- **Not found**
- **Extract unusable**

## Operating

- **Active**
- **Closed**

## Funding Status

- **Public**
- **Private**

## Company Type

- **For-profit**
- **Non-profit**

## Stage And Funding

- **Stage present**
- **Funding observed**
- **Funding obfuscated**

## Scores

- **Growth score**
- **Heat score**
- **CB rank**

## Profile Rollups

- **Competitors listed**
- **Founders present**
- **Employee range**

## Funding depth

Org-page stage ≠ rounds inventory. Prefer this section when you need depth.

## Funding Footprint

- **Financials observed**
- **Not found**
- **Extract unusable**

## Rounds

- **Rounds observed**
- **No rounds listed**

## Investors

- **Investors present**
- **Lead investors**

## Round Recency

Latest round recency

Days since latest parseable round date — extract parse, not verified close.

## Growth & milestones

Trend points are vendor snapshots — not proof of causal growth or buyer intent.

## Growth Footprint

- **Growth page observed**
- **Not found**
- **Extract unusable**

## Trends

- **Growth trend**
- **Heat trend**

## Milestones

- **Milestones present**
- **None listed**
- **Milestone recency**

## Tech & traffic

Pair with Website Ranking when you need an independent popularity ladder — Crunchbase visits are a third-party estimate on the tech page.

## Tech Footprint

- **Tech page observed**
- **Not found**
- **Extract unusable**

## Visits

- **Visits observed**
- **Visits unavailable**
- **Visit growth**

## Traffic trend

- **Accelerating**
- **Stable**
- **Declining**

## Stack Cards

- **Stack observed**
- **Stack empty**

## It Spend

Projected IT spend

USD projected IT spend — estimate only, not budget fact.

## Boundaries & method

- A Crunchbase page is a listing / AI extract — not ICP fit or complete firmographic truth.
- Sparse or unusable extract ≠ the company has no Crunchbase presence.
- Operating active/closed on Crunchbase is a snapshot — not Companies House or legal proof.
- Public / Private are taxonomy labels — not exchange verification or full cap-table secrecy.
- Total funding is USD when coercible; obfuscated means we won’t invent a number — and we don’t invent size bands.
- Growth score, heat score, and rank are vendor snapshots — not buyer intent or Hot/Warm.
- Growth/heat trend points and milestones are proprietary / sample observations — not causal growth proof.
- Rounds and investor lists are inventory rollups — not audited raises or complete rosters.
- Visits, visit growth, and projected IT spend are third-party / projected estimates — not analytics or budget fact.
- Technology counts are a detection sample — not an inventory of record. Profiles / contacts depth is not part of this guide.

---

# Firmographic & technographic

HTML: https://propensify.io/data-sources/firmographic-technographic
Markdown: https://propensify.io/data-sources/firmographic-technographic.md

Industry, type, and HQ are categorical here. Stack language lives on Website signals and Job openings.

- **Also read:** [Website signals](https://propensify.io/data-sources/website-signals) — stack language on their own site
- **Also read:** [Job openings](https://propensify.io/data-sources/job-openings) — tooling and process language in the requisition

## Firmographics bind the account. Tech shows the stack sample.

### Firmographics

### Technographics

## How enrichment evidence stacks

1. **Footprint** — Observed / miss / thin
2. **Firmographic profile** — Type · size · industry · geo
3. **Tech stack** — Names present or unavailable

## Enrichment footprint

Exactly one footprint outcome per research pass.

## Footprint States

- **Profile observed**
- **Domain not found**
- **Thin / unusable**

## Company type

At most one type on a pass.

## Type States

- **Private**
- **Public**
- **Educational**
- **Nonprofit**

## Firmographic profile

Category tags when present on the profile — taxonomy coverage varies.

## Profile Cards

- **Size band**
- **Revenue band**
- **Industry**
- **Country**

## Category Chips

- B2B
- B2C
- SaaS
- E-commerce
- Manufacturing
- Healthcare

## Funding snapshot

This is a snapshot lane — for round-by-round depth, pair with Crunchbase.

## Funding Cards

- **Funding stage**
- **Total funding**
- **Funding rows**

## Tech stack

A name list is not an exhaustive stack audit. Unavailable ≠ no software.

Illustrative names when a stack sample is observed — not every tool the company runs.

## Tech States

- **Stack observed**
- **Stack unavailable**

## Tech Chips

- Shopify
- React
- Google Analytics
- Salesforce
- AWS
- WordPress

## Discovery URLs

## Discovery Cards

- **LinkedIn URL**
- **Crunchbase URL**

## Boundaries & method

- Third-party enrichment for the domain — not registry legal truth.
- Domain not enrichable here ≠ the company is illegitimate.
- A thin return means this pass lacked usable keys — not “no firmographics exist anywhere.”
- Employee range is an enrichment band — not payroll headcount or live workforce count.
- No revenue band on this pass ≠ low revenue.
- Technology names are a sample list — not an exhaustive stack audit.
- No tech names returned ≠ the company uses no software.
- Stage / total / row count are enrichment snapshots — may not match Crunchbase depth.
- Reported LinkedIn / Crunchbase URLs are discovery hints — they can be stale.
- We do not invent size or revenue ladders beyond the enrichment bands, and this guide has no “recent raise” timing.
- Headcount motion and people coverage live on the Company workforce guide.

---

# Job openings

HTML: https://propensify.io/data-sources/job-openings
Markdown: https://propensify.io/data-sources/job-openings.md

## Three questions we ask of the requisition

1. **What function is this, really?**
2. **How is the work done?**
3. **What motion does the copy imply?**

No openings, or a description too thin to read, is unknown — not that they never hire.

- **Also read:** [Company workforce](https://propensify.io/data-sources/company-workforce) — listings ≠ headcount series

## Footprint is a gate. Timing is prioritization.

### Footprint

### Timing

## How job-opening evidence stacks

1. **Footprint** — Active / none + texture
2. **Lifecycle** — New, aging, closing
3. **Function momentum** — Heating → quiet by family

## Hiring footprint

Active and none are exclusive outcomes for a research pass. Stale and closed can appear alongside active listings.

## Footprint States

- **Active openings**
- **None returned**
- **Stale open**
- **Closed listings**

## Listing lifecycle

## Lifecycle texture

- **Newly first seen**
- **Long-running open**
- **Quickly closed**
- **Recently closed**

## Function momentum

Listing momentum is not headcount growth. Closure speed is observed listing lifetime — not confirmed time-to-hire.

## Category motion

- **Accelerating**
- **Recently emerged**
- **Cooling**
- **Dormant**

## Closure Velocity

- **Fast closure**
- **Slow closure**

## Mix & clusters

## Mix Cards

- **Function cluster**
- **Location cluster**
- **Seniority concentration**
- **Open-mix concentration**

## Keyword focus

When your ICP supplies focus roles or keywords, we can match titles, taxonomy, and recent keyword motion — including lower-precision description-only hits.

## Keyword Chips

- Role match
- Title match
- Taxonomy
- Description-only
- Open concentration
- Recent
- Accelerating
- Long-running

## Boundaries & method

- Public / observable openings for the domain — not every ATS on earth.
- Open listings are visibility — not confirmed hires.
- Closed does not prove the role was filled.
- No openings returned ≠ the company never hires.
- Stale open may be evergreen ATS noise or quiet hiring.
- Long-visible listings aren’t proof of a hard-to-fill role.
- Short listing lifetime isn’t a confirmed quick hire.
- Function acceleration is listing observation — not headcount growth.
- Clusters reflect the returned openings sample for the domain.
- Salary when disclosed uses annual USD bounds — not total-comp truth.
- Keyword focus only when your ICP supplies keywords; description-only matches are lower precision.
- We do not invent open-role volume bands (high / medium / low headcount from listing counts).

---

# Company workforce

HTML: https://propensify.io/data-sources/company-workforce
Markdown: https://propensify.io/data-sources/company-workforce.md

We measure mix and trajectory here. Leader names and stories live on Website signals; how work feels is Employee voice.

- **Also read:** [Website signals](https://propensify.io/data-sources/website-signals) — leader names and stories on the team page
- **Also read:** [Employee voice](https://propensify.io/data-sources/employee-voice) — how work feels, not the headcount series

## Presence is a gate. Motion is prioritization.

### Presence

### Motion

## How workforce evidence stacks

1. **Footprint** — Observed / none usable
2. **Trajectory** — Growth → below peak + events
3. **Department motion** — Mix + growth / contraction

## Headcount footprint

Exactly one footprint outcome for this lane. No usable observation is not proof of zero employees.

## Footprint States

- **Headcount observed**
- **No usable observation**

## Headcount trajectory

Observed series change — not a confirmed hiring or layoff program.

Texture can coexist with a primary trajectory.

## Primary trajectory

- **Sustained growth**
- **Sustained contraction**
- **Below prior peak**

## Trajectory texture

- **Plateau / stall**
- **Recovery arc**
- **Oscillation / volatility**

## Period events

A single burst usually needs broader trajectory context. Contractor cycling is a niche proxy, not strategy proof.

## Event Cards

- **Headcount burst**
- **Headcount cliff**
- **False dawn**
- **Band crossing**

## Department mix & motion

## Department Cards

- **Department dominance**
- **Low function share**
- **Department growing**
- **Department contracting**

## People coverage

Page sample under filters — not a full org chart. No matches ≠ no such employees exist.

Illustrative mix texture when matches return — page-local counts, not a census.

## People States

- **People matches**
- **No people returned**
- **Title query hit**
- **Seniority mix**

## People Chips

- Executive
- Manager
- IC
- Engineering
- Sales
- Marketing
- HR
- IT

## Boundaries & method

- Headcount is LinkedIn-visible observation — not confirmed payroll.
- No usable observation ≠ the company has no employees.
- Trajectory movement is observed series change — not a confirmed hiring or layoff program.
- Below a prior observed peak is not proof of distress intent.
- Recovery arc ≠ renewed buying intent.
- Bursts and cliffs need trajectory context; false dawn is observation instability.
- Contractor cycling is a niche proxy — not proof of a contractor strategy.
- Department shares use observed taxonomy — coverage bias possible; low share ≠ function absent.
- People results are a filtered page sample — not a full org census.
- No matches for a filtered search ≠ no such employees exist.
- Title match quality varies — not a confirmed role inventory.
- We do not invent headcount volume bands (large / mid / small) from observed counts.
- Firmographic size bands and public job openings are sibling guides — not substitutes for this series.

---

# Employee voice

HTML: https://propensify.io/data-sources/employee-voice
Markdown: https://propensify.io/data-sources/employee-voice.md

## Three questions we ask of employee language

1. **Where does execution stick?**
2. **Where is GTM unhappy?**
3. **Is the house under strain?**

No reviews in this pass is unknown — not a happy workplace.

## Brands

- Glassdoor
- Indeed

## Workplace reputation vs compensation texture.

### Reputation

### Compensation texture

## How employee voice stacks

1. **Ratings** — Overall / categorical
2. **Reviews & culture** — Themes · DEI
3. **Pay & process** — Salaries · benefits · interviews

## Ratings footprint

## Ratings Lanes

- **Workplace ratings**
- **Ratings thin / missing**

## Reviews & culture

## Reviews Lanes

- **Employee reviews**
- **Culture / DEI**

## Pay & benefits

## Pay Lanes

- **Salaries**
- **Benefits**

## Interviews & FAQ

## Interviews Lanes

- **Interviews**
- **FAQ / Q&A**

## Coverage honesty

## Coverage Lanes

- **Conversations**
- **Workplace photos**
- **Employer about**

Job listings on Glassdoor or Indeed are not re-heroed here. Pair with Job openings for hiring footprint and Company workforce for headcount series.

## Boundaries & method

- Self-selecting / sparse samples ≠ representative workplace census.
- Missing rating or empty FAQ ≠ negative workplace fact.
- Salary figures are directional approximations — not payroll.
- Culture / DEI / conversations / FAQ pages are often empty.
- Employer about narrative ≠ firmographic or registry truth.
- Job listings and hiring counts are out of scope here (→ Job openings).
- Headcount series lives on Company workforce — not this guide.
- Customer reviews ≠ employee reviews (→ Customer reviews).
- Deep Check evidence only — no Signal Catalogue atoms for this family.
- Platform defaults may be UK-biased.

---

# Amazon Marketplace

HTML: https://propensify.io/data-sources/amazon-marketplace
Markdown: https://propensify.io/data-sources/amazon-marketplace.md

Channel identity and attribution live here. Channel voice — returns, defects, stockouts — lives on Customer reviews and Company socials.

- **Also read:** [Customer reviews](https://propensify.io/data-sources/customer-reviews) — returns, defects, felt-tricked language
- **Also read:** [Company socials](https://propensify.io/data-sources/company-socials) — stockout comments and in-channel voice

## Presence is a gate. Timing is prioritization.

### Presence

### Timing

## How marketplace evidence stacks

1. **Reviews** — Demand
2. **Sales rank** — Purchase
3. **Catalogue** — Footprint
4. **Feedback** — Seller activity

## Timing signals

One clear signal can prioritize an account. Stack a few when you want a stronger reason to reach out now.

## Review velocity

_4 intermediate states omitted for compact budget._

- **Accelerating**
- **Missing**

## Sales rank

- **Improving**
- **Flat**
- **Declining**
- **Missing**

## Catalogue footprint

- **Expanding**
- **Flat**
- **Contracting**
- **Missing**

## Seller feedback

- **Momentum**
- **Stale**
- **Missing**

## Checks

## Checks

- **Presence verified**
- **Active ASIN cluster**
- **No active cluster**
- **Quality at scale**

## When signals stack

## Composites

- **Multi-dimensional growth**
- **Rank leads reviews**
- **Reviews lead rank**
- **Expansion with contraction**

## Growth phase

## Phases

- **Accelerating**
- **Expanding**
- **Mixed**
- **Sustaining**

## Boundaries & method

- Timing windows follow available history (commonly ~90 days); thresholds can be tuned per ICP.
- Review velocity is a lagging demand proxy — not direct GMV or unit sales.
- Sales rank is category-relative; absolute meaning depends on category depth.
- Catalogue expansion can reflect rotation or onboarding — not always net business growth.
- Coverage states (missing, stale) are first-class — we prefer “insufficient data” over weak speculation.
- Seller-level timing requires verified attribution; unattributed brand-store products may inform product demand only.

---

# Customer reviews

HTML: https://propensify.io/data-sources/customer-reviews
Markdown: https://propensify.io/data-sources/customer-reviews.md

## Three questions we ask of review language

1. **Who is speaking?**
2. **What broke or delighted them?**
3. **What does the listing prove by existing?**

No profile in this pass is unknown — not “no customers.”

## Brands

- Capterra
- Trustpilot

## Presence is a gate. Texture is prioritization.

### Ratings footprint

### Review texture

## How review evidence stacks

1. **Profile** — Found / thin / not found
2. **Ratings & volume** — Score + review count
3. **Texture** — Distribution · themes · replies

## Capterra footprint

Not found is common for non-software companies — not proof of illegitimacy or zero customers.

## Footprint States

- **Profile found**
- **Extract thin**
- **Not found**

## Ratings & volume

## Ratings Cards

- **Overall rating**
- **Review count**

## Star distribution

## Distribution States

- **Distribution observed**
- **Distribution unavailable**

## Listing texture

## Listing Cards

- **Pricing disclosed**
- **Free trial**
- **Free version**
- **Reviewer demographics**

## Trustpilot texture

Trustpilot evidence is narrative research texture. Absence of a profile is lack of data in this pass — not a negative quality verdict.

## Trustpilot Chips

- **TrustScore & volume**
- **Claimed profile**
- **Praise themes**
- **Complaint themes**

## Boundaries & method

- Self-selecting review samples ≠ product quality truth, market leadership, or ICP fit.
- Capterra not found is common outside software — not illegitimacy or zero customers.
- No rating trajectory, influencer tiers, or Hot/Warm volume bands.
- Pricing string on a listing ≠ FX-normalized contract value.
- Free trial / free version flags are vendor claims.
- Reviewer demographic lists ≠ statistically valid ICP census.
- Alternatives listed ≠ win/loss outcomes.
- Trustpilot themes are Deep Check narrative — not Signal Check catalogue atoms.
- Trustpilot miss ≠ negative quality signal.
- Customer reviews ≠ employee reviews (→ Employee voice).
- Customer reviews ≠ Ad Library social proof.
- G2 is not in this guide.

---

# Company socials

HTML: https://propensify.io/data-sources/company-socials
Markdown: https://propensify.io/data-sources/company-socials.md

## Three questions we ask of the profile and the posts

1. **What identity does the profile assert?**
2. **How do they sell in-channel?**
3. **What is the demand texture?**

No matching public profile in this pass is unknown — not proof they have no account.

## Platforms

- LinkedIn
- Facebook
- Instagram
- YouTube
- X

## Presence is a gate. Activity is prioritization.

### Profile

### Posting & engagement

## How social evidence stacks

1. **Profile** — Found / not found
2. **Audience** — Size snapshot
3. **Posting & engagement** — Recent → stale + reactions

## Profile footprint

Shared gates across LinkedIn, Facebook, Instagram, YouTube, and X. We do not render five separate found/miss walls.

## Footprint States

- **Profile found**
- **Not found**

## Audience size

No invented large / mid / small influencer tiers. Count when present; silence when not.

## Audience Card

Audience size

Numeric followers or subscribers on the observed profile — snapshot only; not a growth trajectory.

## Audience Chips

- **Followers**
- **Subscribers**

## Posting activity

Same activity ideas across networks. X may lack cadence when dated posts are thin.

## Posting continuum

- **Recently active**
- **Steady cadence**
- **Quiet**
- **Stale**

## Posting Coverage

- **Content present**
- **No content returned**
- **Dates insufficient**

## Engagement

Enrichment depth differs by network. Absence of an engagement atom is not proof of zero interest forever.

## Engagement Card

Engagement observed

Reactions, likes, comments, or views on sampled content when the platform returns them — YouTube views sit in this same lane.

## Platform texture

## Platform Texture

- **Funding on profile**
- **Page ads claim**
- **Reels & access**
- **Shorts & views**

## Boundaries & method

- Organic profiles ≠ Ad Library creatives — pair with LinkedIn Ads, Meta Ads, or Google Ads for paid motion.
- Audience size is a snapshot — we do not invent follower growth trajectories.
- No follower or subscriber “influencer tiers.”
- Content sample ≠ full posting history.
- Not found ≠ never present on that network.
- LinkedIn funding on profile ≠ Crunchbase round depth.
- LinkedIn employee fields on social payloads ≠ Company workforce series.
- Facebook page ads-running claim ≠ Meta Ads Library guide.
- Private Instagram ≠ no business.
- X cadence may be unavailable when dated posts are thin.
- Engagement enrichment cost and depth differ by network — silence is coverage, not a verdict.