Quality checker / EXAMPLE DESIGN
A citation alone does not make a claim correct. See how the Quality Checker compares the number, population and source version, with a reason for every item returned.
Working browser demonstration · Fictional sources · Local rules, no live agent
A proposed claim and the original evidence it is supposed to describe.
A discrepancy report showing what matches, what fails and why.
The named reviewer decides whether to correct or issue the work. A clean check is not professional sign-off.
Compare a current claim with a stale 60% headline. Read the arithmetic and population checks separately. Set the input and select “Prepare this result”. Then read the output before recording a fictional review.
Changing an input starts a fresh example.
Read-only comparison; a discrepancy report is never professional sign-off.
PREPARED OUTPUTA discrepancy report showing what matches, what fails and why.
HUMAN REVIEWThe named reviewer decides whether to correct or issue the work. A clean check is not professional sign-off.
Compare a proposed headline with the original table. Inspect the number, sample and permission separately, with a reason for every discrepancy.
Compare a proposed headline with the original table. Inspect the number, sample and permission separately, with a reason for every discrepancy.
Local fictional example · 0 recorded events · Revision 1
These fictional originals belong to the selected case. Compare the output with the source, version and allowed use. Showing a source here does not clear missing evidence or grant wider access.
Synthetic source. The adjustable N is completed responses, a whole number from 80 to 240 (default 120). Exactly 36 answered yes to repeated evidence requests in simulation week 1; N minus 36 answered no. These categories exhaust the sample. Rate = 100 × 36 / N; display one decimal place. This is invented data, not public research.
Use boundary: Fictional aggregate data may be quoted with the sample, period and synthetic label.
Synthetic source. Proposed text: In this fictional week-1 sample, 36 of N respondents (100 × 36 / N percent, rounded to one decimal) reported repeated evidence requests. Retain the denominator and invented-sample qualifier. One numerical claim requires editorial review.
Use boundary: Synthetic fixture; private review within this fictional case only.
Synthetic source. The old headline reads: 60% of advisory firms repeat evidence requests. Its 36 / 60 pilot denominator belongs to a different fictional sample. It is not the current N, and the phrase advisory firms overstates the population. Keep this source only to inspect the discrepancy.
Use boundary: Comparison only; this stale headline is not cleared for publication.
Synthetic source. Use the survey statement, show the no-response category as N minus 36, and ask where repeated requests occur. Do not invent a client story. The proposed channel is an internal example newsletter. Publication, audience consent, delivery, responses and any enquiry attribution have no receipts in this demo; each remains pending its owner.
Use boundary: Synthetic fixture; private review within this fictional case only.
Run the connected case to see what arrives before this step, who reviews its output and what happens when the next owner returns it.
Bring the task, its current owner and where it gets stuck to a no-fee Workflow Fit Call. If a paid System Blueprint is useful, it measures one service line and fully designs three connected workflows. Roles are chosen for those designs. Installation is separately scoped; pricing is quote-specific.
