Screening and watchlists

Screening compares parties against watchlist entries. Because names are transliterated, abbreviated, reordered and misspelled, the comparison is fuzzy — and because a fuzzy comparison produces false positives by construction, the design is arranged so a human sees the evidence rather than the score.

Match, score, decidePermalink to this section

Match on a party

Candidates are found against the party, not against a free-text string, so every hit is anchored to a record with its own identifiers and attributes.

Score on evidence

The score is assembled from what actually corresponded — name similarity, date of birth, nationality, identifier — and each contribution is retained rather than collapsed into a single number.

Decide on corroboration

Adjudication is a human judgement over that evidence. The verdict, the reviewer and the evidence as it stood are recorded together.

List provenancePermalink to this section

The internal list is a separate, editable master; the screening index is a snapshot of it. That separation is what makes a historical hit interpretable: a hit records the list version it was found against, so a review months later can see the entry as it was, not as it has since been amended.

What a hit keepsPermalink to this section

• The score, and the per-attribute evidence it was assembled from.
• The watchlist entry as it stood, so later amendment does not rewrite history.
• The list version, so the state of the source at the time is recoverable.
• The reviewer, the verdict and the time it was reached.