Urgent.News

What's breaking now, across thousands of outlets.

Tech

Our Clean Up Merged Two Builders Who Shared a Yard and Put the Son on Stop for His Father's Debts

Last year we ran a project to remove duplicate customer accounts. Over twenty years our branches had opened new accounts for customers they could not find quickly on the screen, and about six thousand accounts looked like the same customer twice. Duplicates do real harm, from split credit limits to statements showing half the balance. A data quality tool scored pairs of accounts on name, address,…

Last year, a project was undertaken to eliminate duplicate customer accounts from the records of a financial institution. Over the course of two decades, the institution's branches had inadvertently opened new accounts for customers that were difficult to identify on their screens, resulting in approximately six thousand accounts being identified as potential duplicates.

The presence of these duplicates posed significant issues, including split credit limits and statements displaying half the balance. To address this problem, a data quality tool was employed to compare pairs of accounts based on factors such as name, address, phone number, and email, and merge any pair that exceeded a predetermined threshold.

The tool was subsequently manually reviewed for a sample of a few hundred accounts, and the matches were deemed convincing. Consequently, over four thousand and two hundred duplicate accounts were merged over a weekend period. Three weeks after the completion of the merge, a groundworker contacted the credit controller, expressing his frustration.

The groundworker's account had been placed on stop, and the accompanying statement indicated that he was unaware of several invoices. It was later discovered that his father, who runs a building firm from the same yard on the outskirts of a village, had an overdue account. The data quality tool had identified two accounts with nearly identical names and addresses, merging them into a single account.

Upon closer inspection, it was revealed that one hundred and thirty-eight merges had combined different businesses, including fathers and sons, brothers, and firms renting units on the same industrial estate, with their addresses differing only by a unit number, which the tool treated as insignificant. Correcting these errors proved to be a far more challenging task than the initial merging process.

The question that mattered was not the degree of similarity between the accounts, but rather whether they represented the same business. The data quality tool had treated the threshold as a measure of account similarity, rather than assessing whether the accounts belonged to the same business. As a result, it was decided that merges should now be treated as proposals.

Limited companies would only be merged when their registration numbers matched, while sole traders and partnerships would be proposed to the branch that opened the accounts, and someone within that branch would confirm each proposed merge. To ensure reversibility, the original accounts were preserved during every merge. The tool was tested on samples drawn from shared addresses, as these were the areas where the tool was most prone to making mistakes.

Ultimately, the report emphasized that two records that appeared to be alike should be examined by someone familiar with the customers, rather than automatically accepted as a solution to the problem.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in Tech

More from Wednesday 18 November →