Data quality lessons

We found 5,951 jobs posted by a company called "CHECKED." It doesn't exist.

What a quarter-million job postings taught us about data quality

If your careers service is building any dashboard or report using job board data — yours, a vendor's, or a public feed — read this before you trust the numbers.

⚠️ TL;DR One placeholder value sat in the "advertiser" field of 5,951 postings in our dataset — outranking most real employers until we filtered it out. Only ~12% of distinct advertiser names had ever been resolved into a clean, deduplicated company record. 1,756 (advertiser, title) pairs showed 5+ postings landing on the exact same day — 16,341 records sitting inside syndication bursts, not real distinct vacancies.

🚩 Placeholder values hide in plain sight

Values like "Checked" or "Unknown" sat in 5,951 and 1,230 postings respectively. Location/industry strings leaked into the company-name field too. Left unfiltered, a placeholder outranked most real employers.

🪞 The same company, many names

Legal-suffix variants and casing differences fragment one entity into several. Only ~12% of our distinct advertiser text values had ever been resolved into one clean record.

📎 Posting counts ≠ job counts

16,341 posting records sat inside same-day syndication bursts — one role alone showed 59 "postings" within 25 seconds. Raw counts overstate real vacancy numbers.

🕳️ Missing fields aren't random

Salary, postcode, and industry fields are missing at very different rates by source site — a metric that silently drops missing rows can quietly bias what you conclude.

🎯 Why this matters for your reporting
If your office publishes statistics to students, leadership, or accreditation bodies based on job market data — "X% of roles now list salary," "our top placement partners are Y and Z" — these are exactly the quiet distortions that can make a well-intentioned number misleading. Worth a data-quality pass before publication, not after someone asks a hard question about it.

Build on clean data

We do this cleanup work systematically at GetJobzi — happy to share our methodology with institutional data teams.

👉 GetJobzi.com — Talk to us