Screening customers with High Precision Matching (HPM):
accurate, efficient, and customer-friendly

In 2020, we lost a beloved colleague and one of the founders of Human Inference's natural language processing principles, Holger Wandt. No one spoke as passionately and persuasively about the importance of High Precision Matching (HPM) as he did.

Holger was a passionate knowledge-sharer. Therefore, we have rewritten one of his most popular blogs on his favorite topic, hoping to honor his legacy and inspire others.

Screening customers with High Precision Matching (HPM)

Financial institutions are facing an increasingly complex regulatory landscape. As a result, these organizations must conduct customer screening to comply with laws and regulations and to mitigate risk. Many customer screening system providers claim to offer automated CDD processes. Regular screening of customers is critical for risk management, as it helps to identify suspicious or sanctioned individuals and organizations.

The most significant obstacle in screening lies in the quality of the process. It must be efficient in terms of cost, provide a better customer experience, and offer operational benefits. Achieving this is only possible if you “understand” the data you match.

High Precision Matching (HPM) is a method that uses a combination of probabilistic and deterministic techniques to deliver the best results. It incorporates fuzzy logic and knowledge of names, conventions, cultures, and more to enhance accuracy. As a result, it is a highly advanced and effective method for screening.

Here are some essential requirements summarized:

Matching with various lists

Matching with external and internal lists must be possible, even if they contain non-Latin script names such as Arabic, Mandarin Chinese, and Hebrew.

Strong transliteration capabilities enable High Precision Matching to deliver reliable results and aggregate all matching results from multiple lists into one view.

Generating realistic matching scores

In many lists, data are often misspelled, incomplete, or misordered. For example, "Xao Yin Pin" might be written as "Pin Yin Xao". Additionally, aliases, nicknames, and different date notations are often used. Therefore, it is unrealistic to expect a matching score of 100%.

High Precision Matching generates an accuracy score that matches the quality of the compared records without missing the actual match.

Reducing false positives

When the matching tool produces false positives, it requires manual checks to filter out non-matches.

High Precision Matching reduces false positives by up to 90%. This ensures that the actual risk is not ignored, preventing false negatives.