Current committed rows are annual aggregates for 2015, 2016, and 2017. Person-level rows publish only after source review.
Latest article
The first 497 are not the count. They are the receipt.
A careful count does not begin as a headline number. It begins as a row that can survive being questioned. This first public slice explains why The Counted Press is publishing the ledger in layers instead of pretending the unfinished work is finished.
Reader support pays for record pulls, cleaning, corroboration, and correction work. The site is wired for payment links; secure checkout URLs are the next activation step.
A single death pulled from an MDOC FOIA return does not arrive as a count. It arrives as a row that has to earn its place. Scroll to follow it through four stages.
Stage 01
Raw record
One death is pulled from an MDOC public-records (FOIA) return. The row exists, but nothing about it has been checked. Fields are partial, formats are inconsistent, and it carries no claim yet.
Stage 02
Cleaned
Fields are normalized: date, facility, classification. The row becomes legible and comparable to others, but a clean row is still only an assertion. It can survive editing without yet surviving doubt.
Stage 03
Corroborated
Independent sources are attached. The Counted Press will not publish a person-level row below a floor of three corroborating sources and a confidence of 0.60. Below the floor, the row waits.
sources 3 · confidence 0.72
Stage 04
Published
Only now does the row join the running count. The published tally is the same number the page leads with — prose and figure cannot drift, because the count is built one survived row at a time.
RECORD #0497RAW
date2017-08-14
facilityMarquette Branch
classnatural+other
3 sources · conf 0.72
running count496
Method & sources
Source
MDOC public-records (FOIA) return, 2025-Q1
Corroboration floor
3 sources · confidence > 0.60
Public slice committed
2015–2017 · 497 checked deaths
The four stages every record passes through before it is counted.