The Absolute Futility of Google News for Adverse Media Screening

You wouldn’t use a chainsaw for surgery, so why use Google News for compliance? Without the appropriate adverse media tool, results are inaccurate and imprecise with excessive wasted effort.

Greg Pinn

Greg Pinn

July 30th 2020
Share on

In an effort to reduce compliance spend, many AML executives, BSA Officers, and CAMLOs have relied on “free” tools such as Google News, Yahoo! News, or Bing News to perform adverse media screening and investigations.  These “free” tools are far from free, however, and have a substantial downstream cost on the effectiveness and efficiency of AML analysts and teams.

A 2020 survey of AML/KYC professionals on data usage found that 85% of respondents use Google as part of their risk investigations — a shockingly high number considering Google News is not built for compliance investigations.  As Google News states in its Mission Statement:

…Google products are designed to connect you with a broad array of information and perspectives…

That diverse breadth of information and perspectives are very valuable for general news reading, but is not built for the unique usage of customer due diligence, enhanced due diligence, and AML.

Because Google News (and other similar services) focus on general information on current events, reliance on such solutions greatly increases the workload on compliance analysts while missing critical risk information during the customer due diligence process.  Unfortunately, many in compliance leadership only look at data and technology costs when making evaluations on compliance spend and not the overall costs of running an inefficient program: productivity loss, wasted effort, and missed risk. 

A recent investigation of the global cost of compliance identified a significant difference between labor (57%) and technology (40%) costs for AML compliance.  Based on this distinction, improving the productivity of staff can have a disproportionate impact on technology spend.  Compliance executives must focus their efforts on increasing the efficiency of their teams by identifying cost-effective technologies that can reduce personnel costs while ensuring compliance with regulations and mitigating risk entering their business.

Identifying where purpose-built tools can replace free, generic use tools (such as Google News) to mitigate risk, improve productivity, and lower total costs is critical in building an efficient and effective customer due diligence program.

Lack of Focus

Google News is designed to provide users with access to general news information for individual use.  Results returned from a Google News search covers a huge range of topics and with a change made in December 2017, Google News removed the support for complex URLs that enabled compliance teams to create more complex focused searches.  

With Google News compliance teams have no ability to streamline their news searches or focus their results. They must rely on broad name-based searching, resulting in excessive, unfiltered results that include obituaries, advertisements, press releases, business and technology news, pop culture, sports lineups and scores, and a lot more additional noise that wastes analyst time rather than returning relevant results:  predicate offenses such as money laundering, financing of terrorism, human trafficking, fraud, drug trafficking, and violent crime.

As an example, searching for Raymond Smith in Google News results in pages upon pages of results, with the first relevant result appearing 14th in the list:

This may or may not be the Raymond Smith the analyst is looking for, and so they must continue their investigation. It is another 6 results before a different Raymond Smith with relevant AML concern appears.  

This lack of focus on AML makes using this tool for adverse media screening difficult and error-prone.  Compliance analysts are unlikely to find the needle in the haystack of results as they search, and have to go through dozens of results before they can meet even the most basic threshold for sufficient screening.  This results in significant missed risk since internal compliance policies often set a number of results to review per investigation.

Headlines Only

Google News results provide only the headline of the publication in its results, requiring the reader or analyst to open each article to see if there is some risk that is outside of the topic of the headline.  Going back to our Raymond Smith example, our 27th result provides this headline:

With no information in the result connecting the story to the name Raymond Smith, an analyst must open this article to determine potential relevancy.  In this case, the article discusses a man, Raymond Smith, who died in custody after being booked into jail.  Without any details beyond a headline, analysts must read each article in its entirety to ensure no risk is missed.  From the headline, it is impossible to know if Raymond Smith is the investigator, perpetrator of a crime, victim of a crime, or a man who died under unclear circumstances.

Because Google News only returns headlines, each article click requires the entire source page to load.  Depending on the site, this can cause quite a lot of wasted time; time that ads up when the analyst must review dozens or hundreds of articles a day.  This also opens up the financial institution to cybersecurity risks as firewalls must be kept open to enable this traffic.

Irrelevant Articles Fill The Results

Five of the first ten results from Google News for Raymond Smith return obituaries that contain no risk information.  These and other non-material listings cause analysts to need to read each article in its entirety to ensure no predicate offense related risk is missed.

Continuing with the analysis of Raymond Smith, the first 50 results gathered from Google News produce the following:

Only 20% of articles (10 out of the first 50 results) were material and relevant for an AML screen and, as you can see below, were returned in a seemingly random fashion.

Materiality

A common method of constraining analyst workloads is to set a limit for the number of results to screen.  This policy can only work properly if results are ordered intelligently: by the quality of name match and risk-relevancy.  Therefore, a key metric in measuring the quality of any adverse media tool is “first materiality” or where in the result set the first risk-relevant result appears.  Ideally, all results should be material or results should become less material as the analyst continues their work, providing a key point to stop their research for each case.

With Google News, there is no trend in materiality.  For the Raymond Smith search, the first material article was the 14th result with no trend after that.  Relevant results occurred randomly through the 50 results that were reviewed and likely would have continued after this, providing no way for an analyst to make a risk-based decision on where relevancy in results will end.

The above image illustrates the types and order of results for Raymond Smith in Google News going from the first result returned on the left to the 50th result on the right, color-coded as follows:

With the first relevant material article appearing so late in the result set (#14) and continuing without any pattern, a risk-based approach to adverse media screening is not possible with Google News. There is no method for properly determining when sufficient analysis has been completed.

Loss of Productivity

The features of Google News clearly make it an inappropriate tool for AML Adverse Media screening.  Without any ability to prioritize results based on predicate offenses and AML/KYC risk, analysts must go through dozens, if not hundreds, of results to ensure sufficient risk mitigation.  Because each result presents only the headline, the analyst is forced to load each publication, hoping that the information will not be behind a paywall.  

Additionally, without any integrated tools for record-keeping or reporting, relying on Google News forces BSA Officers, CAMLOs, and other AML management to trust that their analysts are performing the necessary checks and remediations.  Google News provides no audit trail for management teams to ensure proper alignment with internal controls. This poses serious problems from a repeatable risk-based approach to screening.

As demonstrated above, remediating the first 50 hits may not even be sufficient when using Google News due to the lack of prioritization of hits.  The work to remediate these hits and identify the 10 relevant results takes almost 2 hours of analyst time making this “free” solution quite costly in analyst time and productivity, not to mention the missed risk that could result in significant fines or losses due to reputational damage.

A Better Way

Existing solutions, both free ones such as Google News and premium services suffer from the issues raised above.  While legacy AML adverse media solutions may solve the issues around audit trails and reporting, they do not solve the fundamental problem of uncovering relevant risk quickly while minimizing noise.

Merlon was created to solve the issues faced in unstructured adverse media investigations by creating an application built on a bedrock of state-of-the-art artificial intelligence (AI) and natural language processing (NLP).  These technologies solve the issues of relevancy that keyword-based solutions cannot solve.  Merlon’s AI is context-aware, ensuring each hit provides relevant risk details that are tied to the target of an investigation.

Using Merlon, the search against Raymond Smith produces far different results from those of Google News.  First of all, Merlon constrains the results, preventing the analyst from needing to review an endless list of results.  For Raymond Smith, Merlon returns 88 relevant results and automatically dismisses 298 non-material results, many of which are the obituaries returned in the Google News search above.  By automatically dismissing these results, the analyst is able to immediately focus their effort on relevant risk.

Merlon’s result set (below), returns relevant results 80% of the time.  Of the irrelevant results returned, 2 are victims of a predicate offense, 4 are an investigation of a death in jail, and 2 are DUI arrests.  

When looking at first materiality (i.e. how soon the first relevant result is listed), Merlon’s first material article appears as the first result, and 18 of the first 20 results are risk-relevant.  

With adverse media screening becoming increasingly important for financial institutions as part of the customer due diligence (CDD) and enhanced due diligence (EDD) programs, financial institutions and regulated businesses can no longer afford to rely on seemingly “free” tools like Google News to perform this critical risk mitigation process.  By investing modest amounts in technology, analysts can be more accurate, efficient, and effective in on-boarding clients.

Individual results also show a huge increase in efficiency with relevant information, risk-flags, and risk excerpts available to the user before they need to open the article. With these features, the analyst can identify relevancy before opening the article, being able to quickly dismiss results where appropriate.

The same investigation that took just under 2 hours to complete with Google News took approximately 10 minutes with Merlon.ai (a 92% improvement).

Real Cost Savings

The average pay for an AML analyst is approximately $25/hr.  Based on an 8-hour workday, an analyst should be able to complete 4-5 Google News investigations of similar complexity to the Raymond Smith example above.  In that same 8 hour day, that same analyst could complete 45-50 investigations with Merlon.

On payroll cost alone, a financial institution can save over $45 per investigation and process the backlog of adverse media investigations in a fraction of the time.  

For 1,000 investigations, investigation time drops from 250 man-days with Google News to 20 man-days with Merlon, enabling a team of 4 to complete 1,000 investigations in a week rather than requiring a team of 50 to do the same work.

Selecting the right tool, based on cutting-edge AI and NLP, ensures institutions minimize regulatory and reputational risk, while focusing effort and reducing total cost.

We built Merlon to allow you to instantly use the power of AI to fight financial crime.