Transition Reporting Review · Open data

Codebook

Every code, its scale, the evidence standard behind it, and the two rules that suppress scores a permissive reading would allow.

← Back to the study
CSV
The dataset
PDF
The report

What is being coded

Each organisation's own published documents, read for what they commit to when they describe a restructuring, closure or transition. Not what the organisation did. What it wrote down.

Documents in scope, per organisation:

Trade press was used only to locate documents and to confirm that an event occurred. No score above zero rests on journalism.


Evidence standard

Every score above zero requires a quotation from the organisation's own document, with a source and a section or page reference. No quotation means the code scores zero.

Where a document could not be obtained at all, the code is recorded as not applicable (blank in the CSV) rather than scored zero. This is why the reduction domain has twenty observations and the other three have twenty-three. A blank means "not established here". A zero means "read the document, it is not in there".


The thirteen codes

Domain A: Reduction

What the organisation says is going.

Code Question Scale
A1 Is the headcount change quantified in the organisation's own document? 0 / 1
A2 Is there a financial figure for the saving, deficit, restructuring cost or quantified funding loss? 0 / 1
A3 Is a completion date given, naming a month, quarter or year? 0 / 1

Reduction % = sum(A) / 3 × 100

Domain B: Transfer

What the organisation says is moving, to whom, with what, and by when.

Code Question Scale
B1 Are the transferred functions named? 0 none · 1 general categories · 2 itemised functions (contracting, grant management, monitoring, safeguarding, finance, technical assurance)
B2 Are the receiving partners identified? 0 not mentioned · 1 generic · 2 named organisations, or a defined class with stated selection criteria
B3 Is the resourcing going to partners quantified? 0 / 1
B4 Is there a transfer timetable distinct from the redundancy or closure timetable? 0 / 1

Transfer % = sum(B) / 6 × 100

Domain C: Residual

What is kept, and what is left behind.

Code Question Scale
C1 Is the retained role stated? 0 / 1
C2 How many residual liabilities are addressed? 0 to 6
C3 Is decision-making authority distinguished from delivery? 0 / 1

C2 counts across six liabilities, one point each:

  1. employment and staff transfer
  2. safeguarding
  3. in-country legal entity and registration
  4. assets and equipment
  5. data and records
  6. donor contract novation

Routine reporting unconnected to the transition does not count. A safeguarding policy statement in the governance section scores nothing; a statement about what happens to safeguarding responsibility on handover scores one.

Residual % = (C1 + C2/6 + C3) / 3 × 100

Domain D: Accountability

Whether anyone could tell if it worked.

Code Question Scale
D1 Are there success measures specific to the transition, with a baseline or target? 0 / 1
D2 Is a review or reporting point stated? 0 / 1
D3 Is the organisation's own public localisation commitment referenced in the transition material? 0 / 1, blank where no commitment is held

General programme indicators score zero on D1. A count of children reached says nothing about whether a handover held.

Accountability % = mean of the scored D codes × 100


Two contestable rules

Both were set before coding and applied to every organisation. Both suppress scores that a permissive reading would allow, and both are stated here so that anyone can recode against the opposite rule.

Dates for developing a strategy are not dates for moving a function. Where a document gives a timetable for producing a future strategy rather than for transferring work, B4 is not met.

A commitment named elsewhere in the report is not a commitment referenced in the transition. Where an organisation names Charter for Change, Pledge for Change or its own localisation strategy in a partnerships section, but not in the material describing its restructuring, D3 is not met. Three organisations would score on a permissive reading: Concern Worldwide, Oxfam GB and World Vision UK.


Reliability

Five organisations were coded twice: once through automated retrieval, once from source text after the documents were obtained directly. The two rounds agreed on 50 of 65 codes, or 77 per cent, against an 80 per cent threshold set in advance. The method failed its own test.

Eleven of the fifteen disagreements are attributable to documents the first round could not read, or to the wrong legal entity being read. The remaining four are figures held in notes the first pass had not reached. None arose from two readers interpreting the same sentence differently.

Access, not judgement, is the binding constraint on this method.


Fields in the CSV

Field Meaning
organisation Legal or commonly used entity name as coded
short_name Label used in the figures
A1D3 Raw code scores. Blank = not applicable, document not obtained
reduction_pct, transfer_pct, residual_pct, accountability_pct Domain scores, 0–100
dedicated_document Whether a standalone transition or partnership document was published
group announced (restructuring publicly reported) or comparison (localisation commitment held, nothing reported)
coding_confidence medium-high or low. Low means retrieval was limited
note One line on what the entry turns on
correction_note Populated after publication where an organisation corrected the record. Blank means no correction received
correction_date Date a correction was applied

Known limits

Three organisations should be revisited rather than relied on:

Crown Agents has three separate entities in liquidation. The entity coded is the one filing under the Crown Agents name with no post-2022 accounts.

This dataset measures what organisations published, in documents written for other purposes, some months after the events they describe. It does not measure whether transitions were handled well, whether power moved, or whether partners were satisfied.


Corrections

Every organisation was sent its coded row and given four weeks to correct the record before publication. Corrections received are marked in the correction_note and correction_date columns.

Corrections after publication are logged in CORRECTIONS.md in this folder and applied to the CSV, with the date of the change. Write to matt@managingtransitions.org.


Citation

Thomson, M. (2026) UK INGO Transition Reporting Review 2026: how UK international NGOs report localisation, handover and programme closure. Managing Transitions, Bristol. Dataset published under CC BY 4.0.

Corrections log

UK INGO Transition Reporting Review 2026. Corrections received and applied.

Every organisation in the sample was sent its coded row, the evidence behind each score, and the quotations attributed to it, four weeks before publication. This file logs every change made to the dataset or the report after that letter went out, whether the correction came from the organisation or from a reader.

Each entry records the date, the organisation, the code or claim affected, what changed, and who raised it. Nothing is removed from this file.


Before publication

No corrections logged yet. Entries go here as right-of-reply replies come in.


After publication

No corrections logged yet.


Corrections to matt@managingtransitions.org. If a score is wrong it gets changed, and the change is recorded here with the date.

← Back to the study