---
title: "State of ATS Ingestion: Auditing 460,000+ Live Tech Roles"
slug: "state-of-ats-ingestion-460k-jobs"
description: "An empirical audit of 460,687 live jobs across 14,814 career targets on 17 ATS platforms, analyzing polling latency, drop rates, and real-time freshness."
published_at: "2026-10-09T08:00:00Z"
updated_at: "2026-10-09T08:00:00Z"
author:
  name: "Pulkit Kumar"
  role: "Lead Systems Architect"
  url: "https://jobs.guaqai.me"
  avatar_url: "/assets/brand/wordmark-light.png"
category: "Engineering & ATS"
tags:
  - "ats-engineering"
  - "greenhouse"
  - "workday"
  - "data-audit"
  - "ingestion"
reading_time_minutes: 8
featured: true
og_image: "https://jobs.guaqai.me/assets/hero-product.jpg"
faqs:
  - question: "How frequently does Employed poll employer career targets?"
    answer: "Target career endpoints are polled continuously in priority tiers, refreshing every 4 to 24 hours depending on historical board posting velocity."
  - question: "Why bypass HTML job boards in favor of direct ATS APIs?"
    answer: "Direct ATS APIs (Greenhouse Harvest, Workday CXS, Lever Postings) eliminate headless browser overhead, prevent Cloudflare anti-bot blocks, and deliver structured metadata directly."
  - question: "Are expired or ghost jobs retained in the index?"
    answer: "No. When an employer closes a requisition or delists a posting, our reconciliation pass marks the listing inactive within one scraping cycle."
cross_links:
  - label: "Explore Live Feed"
    url: "/feed"
  - label: "Public v1 API Documentation"
    url: "/docs"
data_sources:
  - "Employed by Guaq AI Production Index (October 2026)"
  - "Direct Harvest and CXS API telemetry across 14,814 targets"
---

## Executive Summary & Audit Methodology

Over a 24-hour observation window in October 2026, the **Employed by Guaq AI** ingestion engine polled **14,814 verified career endpoints** across 17 distinct Applicant Tracking Systems (ATS). Our telemetry captured **460,687 active postings** and ingested **15,994 newly published positions** within a single calendar day.

Traditional job aggregation platforms suffer from "ghost postings", stale listings, and phantom index bloat because they scrape third-party job boards rather than primary employer endpoints. In contrast, Employed queries first-party employer APIs directly, eliminating intermediary latency and verifying open requisitions directly against the employer's canonical database.

> [!NOTE]
> All figures cited in this report originate directly from our production telemetry database. The data represents live, deduplicated career requisitions verified against live employer ATS platforms.

---

## The 17-Platform ATS Landscape

Modern corporate recruitment relies on specialized ATS infrastructure. Each vendor exposes unique API characteristics, pagination behaviors, and rate-limiting thresholds:

| ATS Platform | Target Count | Active Roles | Verification Method | P95 Latency |
| :--- | :--- | :--- | :--- | :--- |
| **Workday** | 4,120 | 142,500 | Direct CXS JSON API | 840ms |
| **Greenhouse** | 3,890 | 118,200 | Harvest REST API Polling | 210ms |
| **Lever** | 1,840 | 45,100 | Postings Public API | 190ms |
| **Ashby** | 980 | 38,400 | GraphQL & REST Endpoints | 165ms |
| **SmartRecruiters** | 1,420 | 41,200 | V1 Postings Directory | 290ms |
| **Taleo / Oracle** | 810 | 28,900 | Candidate Gateway Scraping | 1,450ms |
| **Jobvite & Others** | 1,754 | 46,387 | REST & Syndication Feeds | 410ms |

### Greenhouse vs. Workday Architectural Nuances

Greenhouse continues to serve as the default ATS for venture-backed technology startups and mid-market innovators. Its public API (`https://boards-api.greenhouse.io/v1/boards/{board_token}/jobs`) returns full job definitions, department taxonomies, and office locations in a single unauthenticated HTTP GET call.

Workday, conversely, powers Fortune 500 enterprises. It utilizes enterprise CXS endpoints (`https://{tenant}.wd1.myworkdayjobs.com/wday/cxs/{tenant}/{career_site}/jobs`) requiring dynamic payload negotiation, POST requests, and 20-item paginated windows.

> [!TIP]
> Ingestion workers querying Workday endpoints implement adaptive backoff and concurrent session reuse to sustain throughput without exceeding tenant connection throttles.

---

## Signal Extraction & Field Inference

Raw job descriptions often conceal critical eligibility parameters inside unstructured text. Employed applies deterministic parsing and hierarchical classification to resolve these attributes:

```json
{
  "title": "Staff Distributed Systems Engineer",
  "company": "Acme Cloud",
  "location": "Bengaluru, India",
  "yoe_min": 6,
  "work_arrangement": "hybrid",
  "visa_supported": true,
  "education_req": "bachelors",
  "tech_stack": ["Go", "Kubernetes", "gRPC", "PostgreSQL"]
}
```

Key classification outcomes across our audited 460k dataset:
- **Work Arrangement Stated**: 42.1% of postings explicitly declare remote, hybrid, or on-site status. For the remaining 57.9%, our geographic heuristics infer placement based on office headquarters.
- **Visa Sponsorship Transparency**: Only 8.4% of postings explicitly state visa sponsorship eligibility in the requisition header. However, deep-text parsing reveals sponsor willingness in an additional 14.2% of engineering postings.
- **Experience Requirement Floors**: 68.3% of postings state an explicit minimum Years of Experience (YOE) threshold, with 0-2 YOE comprising just 12.8% of tech postings.

---

## Frequently Asked Questions

### How often are ATS endpoints polled?
Target career endpoints are polled continuously in priority tiers, refreshing every 4 to 24 hours depending on historical board posting velocity and hiring volume.

### Does Employed charge candidates to apply?
No. Employed indexes open roles and links every role directly to the employer's official application page. Candidates never pay to view or submit applications.

### How does Employed prevent duplicate listings?
Every requisition is fingerprinted using a canonical hash of employer identifier, normalized title, location scope, and internal ATS requisition ID, preventing cross-board duplication.