Workplace Program Detector
Pricing
from $6.80 / 1,000 domain analyzeds
Workplace Program Detector
Detects which people programs a company publishes: employee resource groups, DEI, wellbeing and mental health, learning and tuition support, parental and caregiver leave, volunteering. Reads careers, culture, benefits and ESG pages plus live job postings. One flat row per domain for Clay.
Pricing
from $6.80 / 1,000 domain analyzeds
Rating
0.0
(0)
Developer
Mamba Labs
Maintained by CommunityActor stats
0
Bookmarked
1
Total users
1
Monthly active users
11 hours ago
Last modified
Categories
Features
Share
🔎 What can Workplace Program Detector do?
Give it a company domain and it returns one flat row naming the people programs that company publishes: employee resource groups, DEI, wellbeing and mental health, learning and tuition support, parental and caregiver policy, and volunteering and giving.
Every row also reports how much of the company the actor actually read, so you can tell a real "no program" apart from a page it could not open.
| 📦 What you get | ⚙️ Features and integrations |
|---|---|
| 🏅 Six program families, one boolean each 🔗 Evidence URL and phrase per family, so you can check the finding 📊 Coverage block, what was reached, blocked or thin 🧾 34 flat fields, snake_case, one row per domain | 🌐 Two retrieval paths, website pages and live job postings 🤖 Auto board discovery for Greenhouse, Lever and Ashby 🧊 14 day cache, with a skipCache override⬇️ Export to JSON, CSV, Excel, HTML or XML |
Bought by teams selling into HR, People Ops, benefits and learning and development, and by employer brand and DEI consultancies building prospect lists from what companies publish.
🚫 This is not an employer review or ratings source. It does not score a company, rate its culture, or tell you whether a program works. It records what the company published on its own website and job board, and nothing else.
💡 Why use Workplace Program Detector?
| If you sell | Read these fields |
|---|---|
| ERG and community platforms | has_erg, erg_evidence_url, erg_evidence_phrase |
| DEI programs and consulting | has_dei_program, dei_evidence_url |
| Mental health and wellbeing benefits | has_wellbeing_program, wellbeing_evidence_url |
| Learning platforms and tuition benefits | has_learning_program, learning_evidence_url |
| Family and caregiver benefits | has_parental_policy, parental_evidence_url |
| Volunteering and giving platforms | has_volunteering_program, volunteering_evidence_url |
| Anything, as a disqualifier | program_count, coverage, pages_reached |
🧭 Two retrieval paths, because one path misses half the answer
The actor reads company web pages and live job postings, and it ships both because they find different things.
Measured during the build: across five real job boards covering 1,764 open roles, employee resource groups appeared in zero job bodies. Across seven company websites, employee resource groups appeared on two. Benefits language runs the other way, and shows up in job bodies that marketing pages leave out.
signal_source on every row tells you which path produced the finding.
📋 What data can Workplace Program Detector extract?
38 fields per domain. The ones buyers use:
| Field | What it holds |
|---|---|
has_erg | Company publishes employee resource groups |
has_dei_program | Published DEI program, excluding legal boilerplate |
has_wellbeing_program | Wellbeing or mental health program |
has_learning_program | Learning, development or tuition support |
has_parental_policy | Parental or caregiver policy |
has_volunteering_program | Volunteering or giving program |
program_count | How many of the six were found |
*_evidence_url | The page the finding came from |
*_evidence_phrase | The wording that matched |
signal_source | Which path fired, website or job postings |
coverage | How much of the company was read |
pages_attempted, pages_reached, pages_reached_list | What was tried and what opened |
pages_blocked, pages_thin | Refused pages, and pages with almost no readable text |
ats_provider, ats_slug_used, jobs_scanned | Which job board was read, and how many jobs |
render_mode | client_rendered when the page needs a browser to show text |
fetch_status, fetch_error | ok, blocked, or the error that stopped it |
⚠️
falseandnullare not the same thing.falsemeans the pages were read and the language was not there.nullmeans not enough was read to say. They are never collapsed into each other. A company that returns 403 on every path comes back with all six fieldsnullandfetch_status: "blocked", not six confident falses. Readcoverageandpages_reachedbefore you trust anyfalse.
🛠️ How to find a company's employee programs
- Open the Input tab and put a bare domain in
domain, for examplehubspot.com. - Leave
scan_web_pagesandscan_job_postingsontrueso both paths run. - Click Start.
- Read
program_countand the sixhas_*booleans for the answer. - Read
coverageandpages_reachedbefore you act on anyfalse. - Export from the Output tab, or pull the row through the API.
🧪 Using it in Clay
Add it as an Apify enrichment, map your domain column to domain, and the 38 fields land as
one flat row with no reshaping. Every input is accepted as a string, which is what Clay sends.
Gate the run on your ICP column so you only spend the event on accounts you would actually work.
⚡ Skipping board discovery
If you already know the company's job board slug, put it in ats_slug and the actor reads
that board directly instead of discovering it. That is the fastest path when you are running
a list you have already resolved.
💵 How much does it cost to detect workplace programs?
One domain-analyzed event per domain.
| Plan | Price per domain |
|---|---|
| Free | $0.008 |
| Bronze | $0.0076 |
| Silver | $0.0072 |
| Gold | $0.0068 |
There is also an Actor start event at $0.00005, charged once per run per GB of memory.
💳 A domain that turns out to be unreachable is still billed. The actor did the work of attempting it, and a
nullrow that tells you the site blocked us is a real answer. What is never billed is a run that does not start.
Free plan: this actor sets no monthly result cap, so a Free plan run is limited only by your Apify platform credit.
Your spending limit is honored. The run stops as soon as Apify reports that the maximum cost per run you set is reached, and the optional max_charge_usd input stops it before any charge that would cross your own cap. Every row already charged stays in the dataset, and each row lists what it was charged in billable_events_json.
⌨️ Input
Everything is on the Input tab. The options worth explaining:
| Field | Type | Default | What it does |
|---|---|---|---|
domain | string | required unless domains is set | Bare domain, for example hubspot.com. Protocol and path are stripped. |
ats_slug | string | empty | Skip board discovery and read this Greenhouse, Lever or Ashby slug directly. Ignored on a batch of more than one domain. |
programs | array | all six | Report only these program families. See below. |
scan_job_postings | boolean | true | Read the company's live job bodies. |
scan_web_pages | boolean | true | Probe the careers, culture, benefits, DEI and ESG paths. |
max_pages | string | 14 | Clamped to 1 to 25. Sent as a string for Clay. |
skipCache | string | false | true forces a fresh crawl past the 14 day cache. |
domains | array | empty | A list of domains, one row each in one run. Use it or domain; the list wins when both are set. Each domain is billed as its own row |
max_charge_usd | number | empty | Optional cap in USD for this run. The run stops before any charge that would cross it |
🎯 Reporting on one program family
programs narrows the answer to the families you care about. Six are available:
erg, dei, wellbeing, learning, parental, volunteering.
Two rules the row follows, and they are the same rule:
- A family you did not select reports
null, neverfalse, on its boolean, its evidence URL and its evidence phrase.falsemeans the actor read the corpus and did not find it.nullmeans nobody asked. program_countcounts only the families you selected, so a run narrowed to parental leave reports 1 or 0 rather than a number that includes families you never asked about.
A selected family that is genuinely absent still reports false, because that one is a real
measurement. programs_requested on every row records what was asked for.
Note that narrowing does not make a run cheaper. The corpus is the same set of pages and job bodies either way; only the matching is narrowed.
📤 Output
One row per domain, exportable as JSON, CSV, Excel, HTML or XML. 34 flat snake_case
fields, with null rather than a missing key.
Every row also carries these fields:
| Field | What it holds |
|---|---|
dei_evidence_phrase | The matched phrase plus up to 120 characters of surrounding text, scrubbed the same way as the ERG phrase. |
wellbeing_evidence_phrase | The matched phrase plus up to 120 characters of surrounding text. |
learning_evidence_phrase | The matched phrase plus up to 120 characters of surrounding text. |
parental_evidence_phrase | The matched phrase plus up to 120 characters of surrounding text. |
volunteering_evidence_phrase | The matched phrase plus up to 120 characters of surrounding text. |
run_date | ISO timestamp of the run. |
row_status | One word verdict for the row, always present: ok (the look completed and found data), no_match (the look completed and found nothing), error (the look could not complete, read error_reason), or summary (a notice row, not a result). |
error_reason | Why the row is not a clean result, in plain words. Null on a clean row. Always present as a key. |
💡 Tips
- Run it monthly on the same list and diff two rows. That turns a snapshot into a trend.
- A documented absence is a pitch.
has_wellbeing_program: falsewithcoverage: "high"is a company whose competitors publish a program and it does not. - Use
program_countas a cheap sort. Companies publishing five or six families are already investing in this area and are usually the warmer conversation. - Set
ats_slugwhen you have it. It removes the discovery step.
⚠️ Known limits
A false is only as good as the pages that opened. On a company where one page was
reached, a false means very little. Check pages_reached and coverage on every row.
JavaScript-only careers sites cannot be read. Some large companies serve a careers page
that is an empty shell until a browser runs it. Measured during the build: one site's careers
page held 20 characters of readable text, and another held 1,141 characters inside 84,600
bytes of markup. Those rows say render_mode: client_rendered with pages_thin above zero.
Treat every false on them as unknown.
Some sites refuse outright. One of the seven domains in the build sample returned HTTP 403
on nine of fourteen paths. Those rows come back fetch_status: "blocked" with all six program
fields null.
Some sites ask not to be read. robots.txt is read and honored on every domain before
probing. A site that disallows crawling gets pages_attempted: 0 and null program fields.
That is the site's decision recorded correctly, not a company without programs.
Employee resource groups are almost never named in job postings. If a company blocks its
website, has_erg stays null even when the other families resolve from job postings.
Internal branding defeats keyword matching. Companies name their groups whatever they like. One large company in the build sample calls them Equality Groups and was invisible until that phrase was added. Expect misses on unusual internal naming.
Small companies often have no reachable website. One test target has no working HTTPS
certificate and returns a network error on every path. That is a none coverage row, not a
company without programs.
This is a snapshot, not a trend. The actor does not say whether a program is growing or being wound down. The archival sources that would answer that take between 2 and 18 seconds per query, far more than this actor's whole runtime. Run it monthly and diff instead.
Legal boilerplate is excluded on purpose. "Equal opportunity employer" appears in almost every US job posting and is not counted as a DEI program.
The actor records that a company publishes employee resource groups, and deliberately does
not record which groups. The ERG evidence phrase is capped tight and dropped entirely if
what survives names a protected characteristic. When that happens has_erg stays true and the
evidence URL stays populated, so you keep the finding and its source and lose only the
sentence. The same scrub runs on the DEI phrase.
❓ FAQ
Why is every program field null?
The site blocked the actor or could not be reached. Check fetch_status and pages_blocked.
That row is a coverage problem, not a company without programs.
Does it use a proxy?
No. Every path runs direct with full browser headers. The design answer to a block is an
honest blocked status rather than a workaround.
How fresh is the data?
Results are cached for 14 days per domain and input combination, because careers and benefits
pages change quarterly at best. Pass skipCache: "true" to force a fresh crawl.
Can I run a list instead of one domain?
Yes. Run it per row from Clay, or drive it through the Apify API and collect the dataset.
Does it tell me the names of a company's ERGs?
No, by design. It records that a company publishes employee resource groups and drops any evidence phrase that names a protected characteristic.
🧩 Want other GTM data?
Mamba Labs builds a fleet of enrichment actors that share one flat output convention:
Every actor in the suite takes a domain or a company and returns one flat row, so they stack in the same Clay table without reshaping anything.
🆘 Support
Something wrong, or a company the actor reads incorrectly? Open an issue on the Issues tab with the domain and the row, and we will look at it.
ℹ️ Sourcing and legal. Every field comes from pages the company publishes itself, read directly, with
robots.txthonored on every domain. The row describes what a company published on its own website and job board. It is not an assessment of the company, its workforce, or how well any program works. You are responsible for how you use the output, including under applicable employment and data protection law.
Built by Mamba Labs.
