JobsBoard — Continuous Beta Operations Manual

1. The Continuous Operations Cycle

DEPLOY 
  ↓ 
RUN JOB ACQUISITION 
  ↓ 
MONITOR SOURCES 
  ↓ 
GROW REAL JOB INVENTORY 
  ↓ 
INVITE TESTERS 
  ↓ 
COLLECT FEEDBACK 
  ↓ 
FIX REAL PROBLEMS 
  ↓ 
MEASURE 
  ↓ 
REPEAT

2. Real Job Acquisition Campaigns

Registered & Permitted Live ATS Sources

JobsBoard is configured to ingest directly from authentic ATS endpoints:

Greenhouse: stripe, cloudflare, datadog, figma
Lever: palantir
Ashby: ramp, notion

Running an Acquisition Campaign

cd web
node scripts/acquire-live-sources.js

This script executes genuine HTTPS requests against verified ATS APIs, normalizes URLs to strip tracking tokens, generates SHA-256 content fingerprints to prevent duplicates, updates freshness timestamps, and persists full execution audit logs into SourceExecutionLog.

3. Job Data Quality Scoring System

JobsBoard incorporates a multidimensional Job Quality Score (0–100):

Source Reliability (25%): Verified ATS API adapters / direct employer posts.
Last Verified Freshness (25%): Exponential score decay based on days since last scan or posting.
Description Completeness (15%): Length and semantic depth of job responsibilities/requirements.
Location Quality (15%): Precision of geographic data (city, country or remote classification).
Application URL Validity (20%): Fully formed, valid HTTPS target application link.

Tiers:

90–100: EXCELLENT
75–89: GOOD
55–74: QUESTIONABLE
< 55: LOW_QUALITY

4. Operational Resilience Protocols

Automated Fault Recovery

Scheduler Idempotency: Scheduled ingestion runs are tagged with hourly idempotency tokens (source_run_<sourceId>_<YYYY-MM-DDTHH>), guaranteeing no duplicate simultaneous executions.
Source Anomaly Protection: If an ATS returns 0 jobs due to network or upstream issues, active inventory is untouched (missingScanCount is not incremented).
Three-Scan Expiration Safeguard: A job is transitioned to EXPIRED only after failing to appear across 3 consecutive successful scans.
Degraded Infrastructure Mode: If Redis is offline, rate limiting and queues fall back automatically to in-process sliding window and memory queues without dropping incoming HTTP requests.

5. Phased Tester Cohort Strategy

Cohort	Target Size	Audience	Key Goals
Cohort 1 (Active)	3–5 users	Direct trusted contacts	Friction in registration, broken links, search anomalies, mobile glitches
Cohort 2	10–20 users	Extended colleagues & peers	End-to-end application workflow, salary insights, team management
Cohort 3	50+ users	Friends & family wider release	High concurrency, broad browser diversity, organic feedback volume

Active Cohort 1 Provisioned Invitations

Alice (Seeker): alice.tester@example.com
- Link: http://localhost:3000/register?betaToken=beta_forcrx5Vg_N_vyejgiGPSTDMHnTrmxYC&email=alice.tester%40example.com
Bob (Seeker): bob.tester@example.com
- Link: http://localhost:3000/register?betaToken=beta_DCh9wP-92CiLiGsuBWq2BkjzhNzi77VP&email=bob.tester%40example.com
Carol (Employer): carol.recruiter@example.com
- Link: http://localhost:3000/register?betaToken=beta_iF9jH_sVFLFDWPj1-e5Ar_MrdbKuDSNM&email=carol.recruiter%40example.com

3.4 KiB Raw Blame History Unescape Escape