# Symmetrical 36-cell review protocol and change log

Published 9 September 2026. This document exists because the previous matrix was
built asymmetrically: two observations were coded during targeted searches for
cells that lacked support, and no equivalent search was run against cells that
already had it. That biases a matrix toward being populated. This review applies
one procedure to all 36 cells.

## Protocol

Applied identically to every market and every category, with no exceptions.

1. **Sources.** Each market's full source set as registered, no additions:
   UK SRC-051 to SRC-053; GCC SRC-054 to SRC-056; India SRC-057 to SRC-059;
   US SRC-060 to SRC-063; Singapore SRC-064 to SRC-067; Hong Kong SRC-068 to
   SRC-071.
2. **Retrieval.** The same captures for every market: full JavaScript render,
   headless Chrome, retrieved 9 September 2026, pinned by SHA-256 in
   source-evidence-record.csv. No market was re-retrieved at a different depth or
   duration than another.
3. **Blinding.** Each market reviewer was given the six category definitions and
   the source text, and was not told which cells were previously marked present.
   Reviewers were instructed to search absent categories as hard as present ones
   and to report findings that contradicted an expected pattern.
4. **Evidence standard.** A cell is present only where a verbatim passage of 40
   to 600 characters states the thing, as a sentence or run of sentences.
   Navigation, headings, marketing copy, benefits lists and adjacent material are
   excluded by instruction. Every passage was then verified mechanically as a
   literal substring of the retrieved text, with its character offset recorded.
5. **New observations.** Reviewers recorded every clearly stated selection fact
   they encountered, not only facts that filled a gap. Forty-one were noted and
   are held for coding; none was used to populate a cell in this pass.
6. **Adjudication.** Where the reviewer's call did not survive my own reading,
   the disagreement is recorded below rather than resolved silently in favour of
   the fuller matrix.

## Result

Twenty-six cells. Two rows verified in all six markets: test or structured
assessment, and interview.

| | UK | GCC | India | US | Singapore | Hong Kong | row |
|---|---|---|---|---|---|---|---|
| Prior record or academic threshold | yes | - | yes | yes | yes | yes | 5 |
| Test or structured assessment | yes | yes | yes | yes | yes | yes | 6 |
| Writing assessment | - | yes | yes | - | yes | - | 3 |
| Case, task or simulation | - | yes | - | - | yes | - | 2 |
| Interview | yes | yes | yes | yes | yes | yes | 6 |
| Panel assessment or committee review | - | - | yes | yes | yes | yes | 4 |
| **verified present** | 3 | 4 | 5 | 4 | 6 | 4 | **26** |

## Change log against the previous matrix

| Cell | Before | After | Why |
|---|---|---|---|
| Hong Kong / Panel | not observed | present | Found by the symmetrical pass, not by a targeted search. Freshfields states the candidate is interviewed by a senior lawyer and the people team, which is more than one assessor. |
| Hong Kong / Prior record | present via SIG-106 | present | The same Freshfields academic-record passage was found independently by a blinded reviewer searching all six categories. It no longer rests on a targeted search. |
| US / Panel | present via SIG-107 | present | The Harvard Admissions Committee passage was likewise rediscovered by the blinded pass, and the reviewer additionally located an official-transcript requirement supporting the US prior-record cell on a firmer basis than the graded-paper passage. |
| UK / Interview | present | present | Unchanged. The market reviewer's transcription spanned a line break and failed verification; repaired to the verbatim run and confirmed against the hand-verified passage. |
| Everything else | unchanged | unchanged | |

## Adjudications, where I overruled a reviewer

**GCC / Prior record - reviewer said present, recorded as absent.** PwC states
that "Having the right academic results is only part of it." That describes how
academic results are weighted; it does not state a record, transcript or
threshold used as an input, which is what the category definition requires. The
fuller matrix was the more convenient answer and is not the supported one.

**Hong Kong / Case, task or simulation - reviewer said present, recorded as
absent.** The passage describing "interactive tasks and scenarios" is the
adjacent sentence to the passage supporting the Hong Kong test cell, and both
describe the same Freshfields online assessment. One instrument populating two
cells inflates the matrix. Flagged for founder judgement: if you take the view
that an assessment made of interactive tasks is properly both, the total becomes
27.

## What this does not establish

The result is 26 rather than 25 because a blinded symmetrical pass found one more
cell, not because a target was pursued. It could as easily have fallen. Two of
the 36 cells turn on my judgement rather than on an unambiguous passage, and both
are named above. Nothing here is approved; every ruling remains RECOMMENDED
pending Hassan's signature.