Source classes, freshness and rejections

Freshness and outliers

Rates and salaries older than 18 months, overhead sources older than 24 months, provider facts older than 12 months and FX rates outside the current quarter are rejected. An observation above 2.5 times or below 0.4 times the running weighted median (the middle value of the observations so far) of its combination is excluded as an outlier and stays in the evidence file.

Independence: two sources are independent when publisher and underlying dataset differ; two pages on one domain count as one source.

Rejections by reason

• 38: figure not present in the quoted text
• 22: undated page
• 5: figure is derived, not stated on the page
• 4: older than 12 months
• 3: statutory class requires an official publisher domain
• 1: page shows Oct 18, 2026 for this post (time dateTime="2026-10-18"); Sep 21, 2026 does not appear on the blog index
• 1: page opened but shows no company summary; the quoted sentence, Egypt and incorporation details are not on the page
• 1: page opened but shows no company summary; the quoted sentence and 2005 are not on the page
• 1: page states only the US headquarters; engineering teams are described as operating globally, so a US delivery country is derived, not stated on the page
• 1: starting price without a stated level (inputs/sources.md)
• 1: page could not be read by the verifier (PDF text not extractable); entry has role null and feeds no record
• 1: the source places the role at SFIA levels 1 and 2 (junior); a middle mapping contradicts the stated level