Chapter 25: Nine Red Flags
Seven of the nine red flags began with people asking the company a question.
Friday's data-room screen showed no names, departments, narratives, outcomes, addresses, or contact information.
The WMG custodian had produced a redacted comparison view under Daphne's limited authority and Priya's observer protocol.
Each row contained only a random record ID, an internal event date, a later protocol tag, current-or-former employment status, and no link to a person.
Dana sat beside me with yesterday's D05 page index.
Grant watched from across the table.
“Nine is a review sample, not a population count,” he said.
“Agreed.”
The first row showed an internal complaint number, followed twelve days later by an external stability flag.
The second followed the same order in six days.
By the fourth row, the sequence had become difficult to ignore and still impossible to call causation.
The sample could have been selected because those records matched our request.
A legitimate safety concern could also arise after an internal complaint.
The flag did not tell us what anyone had done to the employee.
“Count the order, not the motive,” Dana said.
We did.
Seven rows began with an internal complaint or protected issue number and later received an external stability tag.
Two did not show that sequence.
One began with a documented worksite threat category.
The other showed only an external media event, with its preceding fields outside our view.
That made the seven a correlation inside a narrow sample, not proof of a companywide retaliation rate.
I asked the custodian what fields existed behind the comparison.
He listed routing time, approving role, status changes, access events, and referral category.
He did not offer identities or private content.
“Request those process fields for these nine IDs,” I told Dana. “Nothing that tells us where a person lives or what the original complaint says.”
Grant asked why nine separate paths were necessary if D05 itself was only a policy.
“Because a written rule cannot prove its own use,” I said. “The paths can show what happened procedurally, not why.”
Dana drafted a request for the nine record IDs, their trigger sequence, approval roles, status history, and system-access trail.
She excluded names, addresses, medical information, family information, complaint narratives, and communication content.
The future independent reviewer, once lawfully selected and authorized, would receive any approved records directly from WMG's custodian.
I would receive neither native files nor identities.
Priya recorded WCT's protection concern without claiming power over the WMG data.
Daphne's board channel received Dana's narrow request.
The custodian logged every row we had viewed and closed the comparison filter.
Before he ended the session, Dana asked whether any record ID could already be matched to a matter number lawfully known to us from prior records.
He displayed only the cross-reference permission column.
One row showed ISSUE 17.
That identifier connected the sample to Jonah without revealing anything new about another person.
His row carried the same order: internal issue first, external stability tag later.
It did not prove the tag caused his firing, the public statement, or any later harm.
Another row carried no matter number we recognized.
Its employment field read ACTIVE.
No name, position, location, or complaint detail appeared.
I could not contact someone I could not identify, and I had no right to demand the identity for my investigation or for public messaging.
The proper question was whether WMG could protect the employee while lawful review authority was still being built.
Dana added an interim non-retaliation and restricted-access request for the active record ID to Daphne's channel.
Grant reserved his objection but agreed the request could be decided without disclosing the person to us.
The custodian set Monday morning as the protection-response deadline.
One of the nine IDs was Jonah's. Another belonged to someone still inside the system.