Adverse media testing

Pilots from Q1 2027

Does your adverse media screening find the right story, about the right person?

Adverse media is the hardest screening to test, and few firms have tried. We're building the test with the same discipline as our sanctions and PEP work, and piloting it with a small group of firms and providers from early 2027.

Why it's hard to test

There's no list to test against

Sanctions and PEPs have official sources. Adverse media truth has to be built story by story.

The same name belongs to thousands of people

Finding a story is easy. Finding it about the right person is the test.

The story may only exist in another language

Local reporting often comes long before, or instead of, English coverage.

Risk is a judgement

An allegation, a charge and a conviction are different things, and categories differ between providers.

What your results look like

See where the misses come from.

The pilot report scores each question separately, so you can see where misses come from: the data, the matching, or the categorisation.

Illustrative · not real results

01 · The right person

Finding the story is half the test.

People with verified adverse stories

Found, right person 71%Merged with a namesake 9%Missed 20%

Same-name decoys

Correctly cleared 83%Flagged with someone else's story 17%

Each wrongly flagged decoy is an analyst investigating the wrong person.

02 · Language reach

Stories reported only in the local language are found far less often.

Also reported in EnglishLocal language only
DE
FR
IT
PL
ES
NL

Share of verified stories found, by market and language of reporting.

03 · Category

Is the story classed correctly?

Fraud88%
Corruption81%
Organised crime72%
Sanctions evasion64%
Terrorist financing58%

A sanctions-evasion story filed under "other" may never reach the analyst who needs it.

04 · Time to detection

How long a new story takes to reach your results.

38%
34%
16%
12%
Within 7 days8 to 30 days31 to 90 daysNot found by next retest

Measured on stories published between one retest and the next.

What the test measures

Six questions, each scored separately.

Designed

Recall

Does it return verified adverse stories about real people, built from reputable reporting and checked by hand?

Designed

The right person

Same-name decoys with no adverse history, or someone else's. Does it tell them apart?

Designed

Category

Is the story classed correctly: fraud, corruption, sanctions evasion, organised crime, terrorist financing?

Designed

Language reach

Stories reported only in the local language, across the markets you agree for the pilot.

Designed

Time to detection

How long a new story takes to reach your results, measured from one retest to the next.

Designed

Noise

Duplicates, low-quality sources and stale stories that cost analysts time without adding risk.

The pilot

A small group, shaping the test with us.

Open to banks, fintechs and payment firms, and to adverse media data and screening providers.

  1. Agree the languages, markets and risk categories that matter to you
  2. Receive a blind set of names, mixed with same-name decoys
  3. Screen it as you normally would and send back what you found
  4. Get your results, and a say in how the method is finalised