Solution 4 of 4 · internal working example · not for circulation

The Artefact Clock

Two clocks. Only one of them is the owner's, and nobody was measuring it.

Stop measuring review time by asking people how long they take. Measure it from the artefact: days from month end to the last-saved stamp on the file, read straight out of the file itself. The charter promises time cut against a baseline measured in week one. That baseline does not exist. This is the instrument that supplies one, and it needs nothing from anybody.

Built from: six packs, one month end, four countries
Read on: 26 July 2026, from file metadata
Anchored corporate received dates: none held
Page: v01 · dry run, 26 July 2026

Prepared by DH analyst · date

Reviewed by second person, never the preparer · date

Released by Aleph corporate · date

Worked example on real files. Every reading below was read out of the artefact itself, from the metadata inside each pack. No reading is reconstructed from testimony and no reading depends on anybody's memory. That is the entire point of the instrument.

Evidence basis

Packs measured, one month end, four countries and one multi-country region

8 to 16

Days after month end at last save, on one month end

Days to the one diarised owner-facing event

Corporate received dates ever obtained

Every date a general manager has given the programme is self-reported and unanchored: close by the 5th, submit by the 6th or 7th, at all four sites interviewed. No transmittal email header has ever been obtained, so not one of those dates can be tested. File metadata does not depend on memory, so it is the only clock reading the programme can take unilaterally. The readings below are class L: tied to a source held independently of the pack's own narrative. They are also a floor, not a measurement of effort, because a last-saved stamp records the last touch by anybody, including corporate opening the file to read it.

Internal working example · properties anonymised · not for circulation

The estate clock, May 2026 month end

Six packs from one month end. Properties are anonymised by region. The reading is days from 31 May to the file's own last-saved stamp.

Artefact	Last saved	Reading	Shape	What the artefact face says
East Africa property, single site	8 June	D+8	39 slides	Cover prints the month. No despatch date.
East Africa property, single site	9 June	D+9	29 slides, 2 native charts, 24 images	Cover prints the month. No despatch date.
East Africa property, single site	13 June	D+13	82 slides, no native charts at all, 71 images	Cover prints the month. No despatch date.
North Africa property, single site	14 June	D+14	48 slides, 37 native charts	Cover names April on a May pack. The file name names May. The two disagree.
North Africa cluster, multi-property	16 June	D+16	31 slides	A version string in the file name reads as 16 June, agreeing with the stamp on the day.
Southern Africa cluster, multi-property	not comparable	–	PDF export	Exported to a fixed format, so it carries no equivalent editing history.

Two clocks, and why the difference matters

The apparent contradiction between what sites report and what the files show dissolves once you accept that these are two different measurements.

Clock	What it measures	What the evidence says	Basis
Clock 1, property to corporate	Month-end close to submission upward.	Reported as the 5th to the 7th at every site interviewed. Not one of those dates has ever been anchored to a document.	Self-reported
Clock 2, owner-facing	Month end to the moment the owner actually receives or sees the month.	The file readings above run to D+16 for a cluster deck. The one owner-facing event the programme holds a date for sits at day 22.	File metadata, plus one diarised event
The gap between them	Where the days actually go.	Roughly a fortnight, and this instrument cannot say where it is spent. A last-saved stamp cannot distinguish a property still working from corporate opening the file to read it.	Not yet measurable

How it is produced

Input: the pack file, exactly as it arrives. Nothing else, and nothing asked of anybody.
Method: read the last-saved stamp out of the file, subtract the month end, and record the date the artefact prints on its own face where it prints one.
Effort: about a day and a half to build, then minutes per pack thereafter.
Second instrument, free: log every edit the sponsor makes to a drafted review, and how long he holds it. That single artefact is simultaneously the house-voice training set, a running measure of the charter's own sniff test, and the only honest baseline available given week one was missed.
Deliberately omitted: a corporate received-date column. It depends on transmittal headers that have never been obtained, and a column that cannot be filled is worse than no column.

What this does not do

A last-saved stamp is not a send date. It records the last touch by anybody, including corporate opening the file. Every reading here is a floor.
It does not measure effort. Sixteen days elapsed is not sixteen days of work, and this page never claims it is.
The owner-facing event is diarised, not observed. It was stated in a recorded interview as happening the following day. DH did not attend and did not confirm it took place.
One month end only. Six packs from a single close is a reading, not a trend. The instrument produces a trend from the second cycle.
The clock says nothing about quality. A fast pack is not a good pack. Time is the charter's third priority, behind insight and accuracy, and this page should never be read as though it were the first.

What it proves

The charter promises corporate review time cut against a baseline measured in week one. Week one produced no clock reading of any kind, and the four site interviews since put none of the timing questions in the interview guide to anybody. Without an instrument, every future claim about time saved is unfalsifiable and the charter's third objective simply cannot be reported. This supplies the baseline from files DH already holds, needs nothing from Aleph, and settles the two-clocks question that four interviews left open: improving the clock the sites report on is invisible to an owner living on the other one.

Notes and abbreviations

Every code used above, expanded. This block appears on every solution page.

Evidence class

L, tied to source.: Checked against something held independently of the pack's own narrative: a trial balance, a ledger export, or the file's own metadata.
P, internally consistent only.: Checked against the pack itself, or against the pack beside it. It means the documents do not contradict each other on this point. It does not mean a figure is right.
U, unverifiable.: Could not be tested, with the reason named: no trial balance, the figure sits inside an image rather than as data, or the comparison file does not exist.

Clock notation

D+n, written D+8, D+9, D+13, D+14, D+16.: Days after month end. D+8 means the file was last saved eight days after the last day of the month it reports, D+16 sixteen days, and so on.
Month end.: The last calendar day of the reporting month. For the readings above, 31 May 2026.
Clock 1.: Property to corporate: month-end close to submission upward. Reported at the 5th to 7th everywhere, anchored nowhere.
Clock 2.: Owner-facing: month end to the point the owner receives or sees the month. This is the clock the owner actually lives on.
Floor.: A reading that can only be too early, never too late. Every date on this page is a floor.

File metadata read on this page

Last-saved stamp.: The date a file was last written, recorded by the software rather than typed by a person.
Native chart.: A chart that is still linked to its numbers, so the figures can be read as data. Contrast an embedded image, which is a picture of a chart and cannot be read, reused or corrected without retyping it.
Version string.: A date-like code inside a file name, typed by a person rather than recorded by the software.
PDF.: Portable Document Format. A fixed-layout export, which is why the one PDF artefact carries no editing history to compare.
Transmittal header.: The date and time stamped on the email that sent a pack. It is the one artefact that would settle clock 1, and none has been obtained.

Terms used above

Sniff test.: The charter's own quality gate: would the sponsor send the drafted review as written, or with light edits, or not at all.
Baseline.: The measurement of how things worked before a change, taken so that improvement can be proved rather than asserted.
Cluster deck.: A pack covering several hotels, consolidated above property level.
BRM.: Business Review Meeting, the monthly pack and review an operator gives a hotel's owner.
Pack.: The monthly document a property produces for its owner. In this estate it is usually a PowerPoint presentation; at one cluster it is an Excel workbook.
DH.: Deep Hospitality.