Chapter 46: Minute Nine
The buyer left at nine minutes and took twenty-four shops with her cursor.
On Tuesday morning, PageTurn opened the grouped retail-listening report inside my clean-room account.
The first exit belonged to an anonymised buyer cohort, not a person I could contact.
No shop name, email address or private comment appeared beside the mark.
Twenty-four prospective shops sat behind that test route if the listening threshold held.
Their participation was provisional; one cursor could not award or remove twenty-four orders.
My coordinator wanted the buyer's reason before lunch.
The privacy rules offered behaviour, not an individual explanation extracted under commercial pressure.
I could see that six more listeners had left between 08:52 and 09:14.
The grouped heat map showed a concentrated decline rather than one impatient cursor. PageTurn released it only after the privacy threshold had been met.
I played the accepted master from minute eight.
Carys had already taken the ledger and lied to her colleague.
Then the narrator repeated the cancellation date, distress-call date and payment date.
The same facts had mattered when they first collided in dialogue. Repetition converted evidence into reassurance after the audience already understood it.
Here they delayed the scratched safe without adding a decision.
The master was locked, and my production order held only four unspent pounds.
Quietly replacing the paragraph would corrupt the buyer comparison and require a new approved version.
Personal funding would still need disclosure, review and a fresh cohort. It could not purchase continuity with yesterday's results.
I refused to disguise a production change as page maintenance.
PageTurn allowed a different test variable: the public preview could end earlier.
The complete fifteen-minute master would remain untouched behind the test record. Its original hash stayed visible throughout the page experiment.
I opened the preview cutting table, which referenced rather than edited the source file.
The existing page streamed all fifteen minutes.
I placed a proposed endpoint at 08:44, after the ledger entered Carys's coat.
The cut preserved the seventh call and the decision that defined her risk.
It withheld the repeated recap and the later safe sequence from free listening.
That meant buyers would judge conversion from a shorter proposition.
It also hid six minutes of paid performance and mixing behind the eventual product. That was a real cost, not an accounting fiction.
The actors' credits remained attached to the full master and the preview page.
The page did not call the shorter version a complete sample.
PageTurn generated a cut-point sheet with source timecodes and no new audio hash.
The accepted master hash remained the one delivered yesterday. The cutting table stored only start, endpoint and page-route identifier.
I signed SHORT PREVIEW TEST — MASTER UNCHANGED.
The platform randomised the original and shorter page routes across a fresh buyer group.
No earlier listener was recycled into the new comparison.
The fifteen-minute route repeated the decline around minute nine.
The 08:44 route reached its declared endpoint without that artificial abandonment point.
Its qualified completion rate rose by eleven percentage points.
The result did not prove the unseen six minutes were worthless. Buyers who continued later would still receive the accepted work.
It proved they were badly placed before a buyer had chosen to continue.
PageTurn stored the anonymous heat map, random assignment and page update log.
My account received only thresholded aggregates and the approved cut-point table.
I could not ask which buyer had left or whether she personally controlled twenty-four shops.
The retail team released the shorter preview route for the next conversion stage.
My fifteen-minute master remained sealed in the vault, exactly as accepted.
At 16:12, the grouped completion panel turned green.
The status restored Saltmere's paper-cost development review without creating a print contract.
Then North Quay's retail asset team opened the next requirement.
PRINT TEST COVER — FINAL FILE DUE WITHIN 48 HOURS.
The audio had survived by showing less; the cover now had to identify me at a glance.