Search a customer’s name and watch what comes back. A namesake’s messy divorce. A one-star product review. An angry forum post from a disgruntled ex-employee. A complaint from 2019 that was quietly dismissed. A critical blog nobody reads. And, buried somewhere in the pile, the one thing that actually matters: a fraud investigation.
Every one of those results is, technically, negative news. Almost none of it is adverse media in the sense that should change a decision. That gap, between “there is negative information about this person” and “there is a reason to worry,” is the entire discipline of adverse media screening.
Which is why the question most people ask about adverse media, “is there any negative news?”, is the wrong one. There is always negative news. For a well-known company or a common name, there is an ocean of it. The question that matters is narrower and much harder: does any of it actually count?
That means that adverse media screening is a classification issue, not a search issue; it’s easy to find negative references. The real task consists in working out which of them are genuine reputational warning signs and which are merely noise. This article deals with how to tell the two apart since it is only by getting this right that a screening programme can be one which protects you rather than one which merely creates work.
What adverse media actually is
Start with the definition, because it already contains the answer. The Wolfsberg Group, whose 2022 guidance is the most widely used industry framework for this, defines negative news as public-domain information that a firm would consider relevant to managing financial-crime risk. The same discipline goes by several names, adverse media, negative news, adverse information, reputational-risk screening, with “adverse media” more common in the UK and Europe and “negative news” more common in North America. The names differ; the idea doesn’t, and the Wolfsberg definition is the one most programmes anchor to.
The operative word in that definition is “relevant.” Adverse media isn’t simply bad news about someone. It’s bad news that bears on financial-crime or reputational risk. That single word does a lot of work, because it draws the line between the ocean of merely negative coverage and the much smaller set of information that should actually affect a decision. A parking ticket is negative news. It is not a red flag. A critical restaurant review is negative news. It is not adverse media. A fraud charge is both. Hold onto that word, relevant, because almost everything that follows is a way of operationalising it. Everything after this is really one idea in different forms: relevance, made practical.
Why it matters, and when
Adverse media screening is valuable since it identifies risks at an early stage; someone who is involved in fraud, corruption, or money laundering usually shows up in the news well before they are added to a sanctions list, and frequently well before any conviction, if a conviction ever occurs. It is through checking the public record that you find out about a problem while it is still just a report and not yet a headline or a loss. By the time an issue gets onto a watchlist or reaches a courtroom, the opportunity to act quietly and early has generally gone.
It is also an expectation, not merely a matter of good practice. Although the FATF does not employ the term “negative news screening”, its guidance on a risk-based approach refers to adverse-media searches as part of enhanced due diligence, and supervisors interpret Recommendation 10 as meaning that such screening should be expected in the case of higher-risk relationships. In the United States, FinCEN’s rules on customer due diligence include the requirement for continuous monitoring, and the FCA considers this type of scrutiny to be part of effective financial-crime controls. The common feature is that it is based on risk and proportionate: more intensive in the case of higher-risk customers and relationships and less so in the case of lower-risk ones, rather than applying the same general and indiscriminate approach to all customers. Proportionality applies in both directions: failing to carry out adequate screening of a high-risk customer is a mistake, but conducting excessive screening of a low-risk one simply overwhelms your team with noise. The key is to direct the effort, not to maximise it.
Where adverse media sits alongside sanctions and PEP checks
It is useful to understand how adverse media relates to the other screenings that are already carried out, since each type of screening has a different role. Sanctions and watchlist screening involve checking a name against official, structured lists, while PEP screening determines whether a person holds a position which increases their risk. Both kinds of screening are essential and, by their nature, are lagging indicators: a name ends up on a sanctions list after a government has taken action, and usually long after the actual behaviour in question.
Adverse media is the key early warning sign since it picks up the reporting, the investigation, the allegations and the pattern before any of this becomes a listed entry or a conviction, if it ever does. A person might be deeply involved in fraud or corruption for years and still not appear on any list at all, the only early indication being in the public record. This is the gap that adverse media fills: it identifies the risk that the structured lists are unable to see at this stage. The three methods work together, with the lists dealing with what is officially recorded, PEP focusing on position, and adverse media covering all the actual cases that have not yet been formalised, and a programme which runs the lists but omits adverse media is only monitoring the risk that someone else has already confirmed. In the majority of cases where things go wrong, it is the news that finds out first.
What actually counts: the categories
If the criterion is how relevant a matter is to financial-crime risk, then it is useful to be aware of what meets that criterion. In reality, the categories that are important are those connected with financial crime and the offences which precede it—namely money laundering, fraud, bribery and corruption, evasion of sanctions, terrorist financing, organised crime, trafficking, tax crime, cybercrime, and environmental crime—as well as the regulatory and criminal enforcement that comes after them.
All the other things—such as the divorce, the negative review, the business dispute, and the unflattering profile—count as negative news but would generally not be considered adverse media since they have no bearing on financial-crime risk. There are exceptions and the situation has to be taken into account: a lawsuit claiming fraud is relevant, but one concerning a late delivery is not. Nevertheless, the category filter carries out most of the initial filtering; if a piece of coverage doesn’t fall within one of the risk categories you’re interested in, it almost never ends up in the pile at all. That is also the reason why a robust programme establishes its risk categories in advance through a written taxonomy rather than allowing each analyst to decide on the spot what should be included. It is precisely because of the written taxonomy that you are able to explain to a regulator why you have regarded one story as relevant and another as noise.
The filters that separate signal from noise
Category comes first, not last. Although you are within the correct categories, the majority of the results are still not genuine red flags, and it is five further filters that carry out the sorting. A defensible programme must apply all of them in a consistent manner. If you omit any one of them, a certain type of noise will get through: if you skip identity, you will be overwhelmed by namesakes; if you omit source credibility, rumour begins to count as evidence.
- About relevance and materiality: does this particular item in fact have anything to do with the risk, or is it merely a passing reference, a quote, or a narrative in which the victim is the subject rather than the person who has done the wrong?
- The credibility of the source is an important factor. A report issued by a reputable news agency and subject to editorial control is quite different from an entry on an anonymous forum or a social media statement that has not been verified. The Wolfsberg guidelines make it clear that source credibility is central, so you should give weight to reliable, independent and verified reporting and disregard all the rest.
- This is about identity screening, not just name screening. The most frequent case of a false match is when the wrong person is identified because they have the same name; therefore, verification must go beyond just the name and include the date of birth, location, role, and the other entities associated with the person to make sure that it is actually the individual in question.
- The situation is that an allegation is not an investigation, nor is it a charge or a conviction, and a dismissal or an acquittal should by no means be regarded as a warning signal; it is the position of the case on that scale that determines how much weight it should be given.
- The more distant the origin of an event and whether or not it was eventually resolved, the different its relevance becomes. A case which is still under investigation and one that was settled ten years ago are not the same thing.
Hygiene includes a sixth element: de-duplication. If a single event is reported in fifty different outlets it constitutes one risk and not fifty, and a programme which counts the coverage rather than the event will be overwhelmed by the noise it generates.
Severity isn’t binary: the event-stage spectrum
The event-stage filter should have its own section since it is the point at which effective screening differs from poor screening and where many programmes lose their credibility.
A genuine red flag isn’t merely something “bad”; it has to have a stage, credibility, relevance, and a certain age. The Wolfsberg framework clearly sets out these stages: allegation, investigation, arrest, charge, trial, conviction, and on the other side, dismissal, acquittal, settlement, and regulatory action. The company has to decide at what stage it wishes to be alerted, and that decision is important. Whether it is a well-established fraud conviction reported in a national newspaper or an anonymous claim made on a message board, both cases can be broadly described as adverse, and it is by treating them as equivalent that a screening programme ends up either being paralysed by caution or dismissed as having cried wolf.
A good classification should focus on the gradient rather than giving a simple yes or no answer. Severity, credibility, recency, and stage are four separate controls, not a single switch, and a human can only make a reasonable judgment if they have the ability to see all four aspects. The information should tell you not only that there is something but also how serious it is, how credible, how recent it is, and how far along the process it is, since these are the facts that actually enable a human to make a proportionate decision. If you reduce all of that to a single flag, then you’re discarding the very information that allows someone to decide what action to take.
Three hits, one red flag
Put the filters together and watch them work on a single realistic case. You screen a prospective customer, and the search returns three negative results.
The first is a lawsuit. Read it, and your customer is the plaintiff, not the defendant, suing a former supplier over a contract. Negative in tone, irrelevant to financial-crime risk, and it’s your customer bringing the claim rather than answering one. Filtered out on relevance.
The second case involves a claim of fraud, which seems serious at first but changes when you look at two points: the person mentioned has the same name as the customer but was born on a different date and lives in a different country, and the only evidence comes from an anonymous entry on a complaints forum with no other sources to back it up. It’s the wrong person and the source is weak. The claim has therefore been rejected on grounds of identity and credibility.
The third is a short report, in a national newspaper, that your customer’s company is under investigation by a financial regulator for suspected money laundering. Relevant category, credible source, correct identity, current, and at the investigation stage rather than a rumour. That’s the one. It doesn’t automatically end the relationship, but it’s a genuine red flag a human needs to weigh, and it’s the single item of the three that should ever have reached them.
Three hits, one red flag. The other two weren’t failures of the search; the search did its job by surfacing them. They were the filters doing theirs by clearing them. That ratio, many in and few out, is what a working programme looks like, not a sign that something slipped through. Count the clears, not just the hits.
Why this is hard to do at scale
On a single name, a careful analyst can apply all of this by hand. Across a whole customer base, continuously, it collapses, and understanding why is the key to understanding what a real solution has to do.
The amount of raw data is huge — millions of articles in dozens of languages are published each day. When common names are used, matches will appear that aren’t actually referring to the person in question. Because of transliteration, one name can be spelled in twelve different ways. The same event is reproduced in hundreds of different outlets. Reliable reporting is mixed in with social media commentary and unverified assertions. If you use simple keyword matching, you’ll receive a large number of false positives; if you try to deal with it by hand, you won’t be able to keep up and certainly can’t do it continuously. This is the same kind of noise problem that makes much of the screening so difficult, and the solution is the same: apply the filters mentioned above consistently and on a large scale so that what gets to a human being has already been classified, deduplicated, and assigned to the correct person. The aim has never been to show a human being everything; it is rather to show them only the few items that have survived the filtering process, together with the relevant context.
What you’re actually filtering out
It is important to give the noise a direct name, since the term “false positives” remains an abstraction until you see what it actually consists of. In the case of adverse media, the noise takes on a number of recurring forms.
The main source of incorrect matches is cases where someone has the same name as the person in question, particularly with common names. Duplicate entries, in which a single event is reported by dozens of different outlets, increase the number of entries without adding any new information. Positive and neutral references are included too because a term such as “fraud” appears in a sentence discussing the prevention of fraud or winning an award against fraud. Old cases which have been withdrawn or dismissed continue to reappear. Matters that are irrelevant—such as divorces and negative reviews—remove items from the keyword list but still fail the relevance test. Finally, low-credibility sources, like anonymous postings and uncorroborated comments from social media, appear as if they are reports even though they lack the substance thereof.
Add those together and you have the ninety-plus percent of hits that aren’t red flags. A programme’s real work was never generating that pile, because anything can do that. It’s clearing the pile, accurately and defensibly, so the few genuine signals aren’t lost inside it. A programme judged on how much it surfaces is measuring the wrong thing; the number that matters is how much it correctly clears.
Risk doesn’t only happen in English
One blind spot deserves singling out, because it’s both common and serious: language. A customer’s most relevant adverse media is often in the language of the place they operate, not in English, and an English-only screening programme simply doesn’t see it.
The regional newspapers give extensive coverage to a fraud case, while a regulator’s enforcement notice is published only in the local language, and although a corruption investigation is reported by news agencies overseas, none of these appear when you search using only English. That is precisely why the Wolfsberg guidance considers multilingual coverage to be an essential element of a proper programme. Transliteration worsens the problem since a name taken from another script can be spelled in a number of different ways, and a search which doesn’t take the various spellings into account simply fails to find the relevant results. For any customer who has exposure across borders, carrying out screening that only looks at the English-language press is equivalent to screening with one eye shut, and the eye that is shut is typically directed towards the most high-risk jurisdictions.
What good adverse media screening looks like
Pulling it together, a programme that actually separates signal from noise tends to share a few traits.
- It screens credible sources, and weights them by credibility rather than treating every mention equally.
- It classifies findings by category and relevance, so financial-crime risk is distinguished from merely negative coverage.
- It resolves identity to the right person, instead of flagging everyone who shares a name.
- It records both the stage and the severity of an event, which means that an allegation and a conviction are not regarded as being the same.
- It removes duplicates, meaning that each event corresponds to a single signal.
- All the findings are linked to their sources and can be checked, allowing a person to verify them and a regulator to see the grounds for any decision.
- It keeps going on like this since most adverse media occurs after onboarding and not during onboarding.
Nothing in this list is unusual. Taken together, they represent the difference between a tool that informs you of negative news—which you already knew—and one that tells you whether or not any of it is significant, since that is the only thing you really needed. The first one increases the amount of work for you; the second reduces it, and only the second is worth paying for. The true test of any adverse media tool is whether it leaves your team with less or more to read.
From negative news to a red flag you can act on
The difficulty involved in dealing with adverse media didn’t lie in spotting negative references since there will always be some negative ones; the challenge was in distinguishing the small number that actually involve real risk from the vast majority that don’t, on a consistent and large scale, in a manner that can be defended later on. If you handle this properly, adverse media ceases to cause anxiety and begins to function like an early-warning system.
That’s what Beady is built to do. Its adverse media screening scans more than 100,000 public sources and strips out roughly 95% of the noise, the wrong-person matches, the stale items, the positive coverage, the duplicates, and the merely-negative-but-irrelevant, leaving only financial-crime-relevant red flags, each classified by category and severity and linked back to its original source. It turns negative news, which is everywhere, into adverse media you can actually act on. Less to read, and more reason to trust what is left.
If you want to see what it surfaces on a name you’re weighing right now, and, just as usefully, what it filters out, book a demo and we’ll run it with you.
