Updated October 14, 2026
Every field on MedSales Radar is read from the employer’s own job posting. This page answers the two questions a data buyer should ask any provider: how often each field is stated, measured over every listing on the board right now, and how often we read it right, measured on a random sample checked against the postings themselves.
A blank means the posting does not say. We leave it blank rather than guess, so a lower share describes what employers publish, not a reading we missed. How often we miss something a posting does say is in the next table.
The other measures a buyer tests a data provider on: how much of the market is read, how soon a new posting appears, and how often one requisition shows twice. All three are read live, like the table above.
Measured where the employer’s hiring system dates a posting exactly, over postings dated in the last 30 days at employers we were already reading. That date is a calendar day, so a posting on the board the same day counts as no wait at all.
A random sample of listings, drawn with a fixed seed from the board as it stood on the date shown, so anyone with the same copy draws the same listings.
A reviewer read each posting in full and judged every field against the written rules: right, wrong (including a blank where the posting plainly states the thing), or unclear where the posting is ambiguous or the rule is a deliberate call. Unclear values are shown, not dropped.
Where a posting names no care setting, the call point is an assumption, from what the role sells or the usual setting for a drug rep, and the job page labels it as assumed. A labeled assumption counts as right unless the posting says otherwise.
The reviewers were AI models working apart from the code that fills the fields, and every value they marked wrong was read again before this table was published. Read right is right divided by checked; the range under it is the 95% margin of error for a sample that size (the Wilson interval).
Read from the source: Each employer’s own careers page or hiring system, twice a day. Nothing comes from other job boards or recruiting agencies.
Closed when the employer closes it: Where the hiring system answers, as Workday does, a listing is confirmed closed before it comes off; elsewhere it comes off after three days unseen.
Rules written down: How an industry, a call point, a product line, pay and experience are read is public, on the about page and in the guides, and the reviewers on this page judged against the same rules.
Fixes reach old listings: When a reading rule changes, the postings already stored are read again, so a fix is not only for new listings.
A second reader: A separate model reads postings in bulk and flags where it disagrees with the site. Each disagreement is reviewed before anything changes.
Corrections in public: Errors that readers or our own checks find are listed on the corrections page with the date they were fixed.
History starts in August 2026. Earlier months are not backfilled, so trends cover the weeks since.
A posting is not a hire. A posting taken down may have been filled, paused or moved; we count postings, not headcount.
Only what employers publish. Pay, travel and requirements appear only where the posting states them.
US roles only, from employers whose careers sites we can read; an employer we do not read is not counted at all.
A sample is a sample. The table gives the number checked beside each figure so its precision is plain.