Styxis
Company
Knowledge

Agree the crosswalk before reusing IATI data across TOSSD and CRS

Development-finance reuse needs shared activity, transaction and flow mappings rather than copying similarly named amounts.

Reuse is not copying

Publishing development-finance data once for several reporting purposes requires more than copying an amount with a familiar name. Teams need to understand how activities, transactions, countries and flow classifications correspond. Truthound can check references to an agreed crosswalk; it cannot decide that reporting concepts are equivalent.

The stage of the October discussions

IATI's October 31, 2026 account describes interoperability work with TOSSD and the OECD's CRS at an October 8 Paris forum and October 13–14 Lima assembly. Discussions explored reuse and validation links. They were not an announcement that a new common reporting standard had already been adopted.

A minimal shared crosswalk

A minimal crosswalk should identify source system and key, target system and key, applicability period, relationship type, evidence and owner. Separate activity-level relationships from transaction-level ones. Copying an activity total into each of its transactions can turn reuse into duplicate reporting.

Do not turn 1:m into an error

1:1 is not the only valid relationship. Different aggregation units may require 1:m mappings. Declare the relationship and allocation basis before validation so a legitimate split is not removed as an error. Conversely, allowing m:m says nothing about whether a financial total is preserved.

The reference boundary in Truthound

Truthound Core Python SDK 3.3.4, operating on a prepared Polars table, provides ForeignKeyValidator for supplied child/parent keys and ReferentialIntegrityValidator for caller-provided reference values. A crosswalk owner must establish its scope and keys. Whether one development-finance classification means the same as another requires separate definitional review.

Keep the existing pandas join

pandas 3.1.1 merge(validate=...) can retain 1:1, 1:m or m:1 key-uniqueness checks. Keep an existing join while adding approved-reference membership and unmatched-record review where needed. Share the limitation that m:m validation permits the relationship without checking its appropriateness. A pandas null-key match is not an approved classification mapping; keep unidentified activities outside confirmed crosswalk links.

Unmapped records are shared knowledge

Do not force an unmapped record into the nearest-looking category. Distinguish a definition difference, insufficient evidence and a genuinely inapplicable classification. Assign the next question to its owner. That list is useful shared knowledge about where cooperation is needed, rather than merely a failure count.

Propagate definition changes explicitly

When a definition or identifier changes, identify which crosswalk version is affected. Preserve the mapping used for an earlier report instead of silently reinterpreting it with the new version. Repeated reuse becomes maintainable when a team can explain what a revision connects differently.

The next question about reporting burden

Whether reporting burden falls needs later operating evidence. The immediate artifact can show what is reusable, what remains missing and where totals diverge. When teams can explain those boundaries together, they have a defensible basis for extending automation.

Conditions

Owners supply source and target identifiers and allowed relationships after checking the official definitions.

Alternatives

Retain pandas 3.1.1 merge-cardinality checks while checking membership of the approved crosswalk separately.

Limitations

  • Valid references do not establish identical reporting definitions or demonstrated reductions in reporting burden.

Evidence

Event checked
Subject checked
Modified