Verifex Research
Measured studies of sanctions and PEP screening.
What screening engines really get right and wrong, what is inside the lists, and how to tell. Every report states its method, its uncertainty and its limits.
Reports
VR-2026-06Screening accuracy·
0 of 7
pre-registered gates passed by the engine deployed on October 13, 2026, every failure published
Measuring a sanctions screening engine through its live API, failures included
A pre-registered black-box benchmark of Verifex's live screening API: listed names, altered spellings, other scripts and clean names, with every gate and interval.
Read the report · 20 min readVR-2026-05Sanctions lists·
19%
of UN Security Council entries we did not find on any OFAC list
OFAC lists most new UN targets first, yet about one in five UN entries is not on its lists
We matched all 1,010 UN Security Council list entries to OFAC, EU and UK lists. 19% were not found on OFAC's lists; the UK list has all.
Read the report · 15 min readVR-2026-04Sanctions lists·
28
new SDN listings unscreened on an average day since 2022, weekly runs
OFAC changed the SDN list in 94% of weeks since 2022: what that means for rescreening
OFAC changed the SDN list in 234 of 250 weeks since 2022. A weekly rescreen leaves about 28 new listings unchecked on an average day.
Read the report · 13 min readVR-2026-03Sanctions lists·
61%
of SDN entries were first published on or after 1 January 2022
Six in ten OFAC SDN entries were added after 2021, and nearly half have no alias
61% of the 19,363 OFAC SDN entries were added since 2022 and 48.5% carry no alias. Counts by type, programme, alias and identifier.
Read the report · 14 min readVR-2026-02PEP data·
79%
of removed PEP rows were people who had died
Four in five: what a Wikidata-derived PEP list really contains
We rebuilt a PEP dataset of about 930,000 rows from the Wikidata dump. About 79% of the rows the new definition removed were people who had died. What that means for PEP screening.
Read the report · 12 min readVR-2026-01Screening accuracy·
98.1%
of 2,000 published OFAC aliases retrieved after two fixes (first run: 83.9%)
Real aliases, not synthetic variants: measuring name screening on 2,000 OFAC aliases
An alias-lookup test: 2,000 aliases OFAC publishes, screened against the OFAC list. Recall was 83.9%, not the 92.4% synthetic tests claimed. Two fixes took it to 98.1%.
Read the report · 12 min read
How we do research.
Counts and intervals
Every result is a count with its denominator, never a bare percentage. Results from samples carry a 95% confidence interval.
Methods you can check
Each report says how the data was collected, what was measured and what was left out.
Limits stated first
Self-administered studies say so. Rounded numbers say “about”. Negative results are published.
Data behind every chart
Each figure has its data table one click away, so nothing depends on reading a colour.
Research is written by the Verifex Research team. Self-administered studies are evidence about the cases measured, not independent certification. For plain guides, see the blog.