How to scrape freelance jobs and marketplace demand

Choose the decision before the marketplace

A broad marketplace export is rarely the final product. A freelancer may need fresh projects, a training provider may need skills demand, and a staffing team may need contract roles across sources. The paid unit should match that decision.

For public employment listings rather than marketplace projects, use LinkedIn Contract Jobs Intelligence or LinkedIn New Jobs Monitor.

Run one bounded public sample

1. Open Freelancer Active Projects Scraper and sign in to Apify. No Freelancer login or API token is needed for this Console workflow.
2. Use the input below for one keyword and at most ten projects. Review the input and live Pricing tab before clicking Start once.
3. Wait for that run to reach a terminal status. Continue only after SUCCEEDED; success can still produce zero matching rows.
4. Open the Dataset and verify projectId, title, skills, budget range, bid count, projectUrl, and collection time against the source.
5. Read the free RUN_SUMMARY. An error, partial fetch, or zero-result diagnostic is not a paid market signal.

{
  "keywords": ["python"],
  "skills": [],
  "projectType": "any",
  "maxProjects": 10
}

Ten returned complete projects have a result charge of $0.15 at the verified rate, plus any applicable Apify platform usage. A result limit is not a complete spending cap. Export a useful Dataset to CSV or JSON, then save the reviewed input as your own Task. A public Task is a saved input, not proof of external demand.

Active projects: preserve the source record

The active-project workflow retains projectId, title, projectType, skills, budgetMinUsd, budgetMaxUsd, bidCount, projectUrl, and submittedAt. That is enough to deduplicate a record, revisit the public source, and distinguish a stated range from a later benchmark.

Do not treat a high maximum budget as expected earnings. Keep currency normalization, fixed versus hourly project type, missing values, and the age of the listing visible.

New-project monitoring: baseline first

A monitor needs memory. Its first successful run stores a baseline and should not bill it as a new event. Later runs compare stable project IDs and emit only supported additions. An unchanged check, failed fetch, or prior-hash match remains non-billable.

The monitor outputs eventType, projectId, project details, projectUrl, and detectedAt. The detection time is when the monitor saw the record, not proof of the marketplace's original publication time.

Skills demand: keep the denominator

The skills workflow groups complete project rows and returns skill, category, projectCount, projectShare, medianBudgetUsd, averageBidCount, and sampleProjectUrls. Always preserve the sampled project count, filters, and collection window.

A skill's project share describes the observed sample. It is not total-market demand, freelancer supply, or a forecast.

Budget benchmarks: show sample size and spread

The benchmark keeps projectType, sampleSize, minimumBudgetUsd, medianBudgetUsd, maximumBudgetUsd, medianBidCount, and sampleProjectUrls. Use the median for a resistant centre and the full range to expose spread.

Do not merge hourly and fixed-price work into one headline number. Do not convert missing budgets into zero. A small sample should remain visibly small.

Bid competition: compare like with like

The competition workflow returns skill, projectCount, medianBidCount, averageBidCount, competitionTier, medianBudgetUsd, and sampleProjectUrls. Compare segments collected with the same filters and window.

Lower observed bids can mean lower competition, a newer listing, a narrow skill, or an unattractive project. The tier is a bounded classification supported by the sample, not a guarantee of winning work.

Why there is no broad Neuton Upwork scraper

Upwork provides an OAuth API, but its official terms restrict credential transfer, content resale, and charging users without express permission. Anonymous search also returned HTTP 403. Neuton has not published an unlicensed clone or hidden a login, cookie, proxy, or provider requirement.

Fiverr and PeoplePerHour explicitly prohibit scraping in their official rules. Contra prohibits crawler and data-mining access plus commercial reuse, while Guru's public jobs sitemap does not grant commercial redistribution rights. These products remain blocked unless the marketplace gives written permission for the exact Actor and pricing model.

Common first-time questions