Identify and remove closed businesses from customer's dataset – at scale
A company with data about points of interest regularly removes closed businesses from their dataset to improve accuracy and relevancy. Flagging a business incorrectly – open when it’s closed, or vice versa – makes for a bad user experience and erodes trust in their product. The data team found they were miscategorizing thousands of businesses, so they looked to improve the quality of their data inputs.
They’d been extrapolating from representative samples of credit card data and verifying business closures manually, via web indexing and user input. Plus, they were investing significant time and resources to conduct manual research – making calls and visiting websites to confirm operating status by a set of parameters to define open and closed.
Timely, reliable signals of business health would help the team improve their models and have confidence they were flagging more businesses accurately, much more quickly and efficiently at scale.
Timely signals of business health improve data accuracy
The team piloted Enigma’s Transaction Stability data attribute – unique because it captures actual credit card swipes at a business versus representative samples.
During the pilot, the team evaluated the operating status of businesses in two cities, comparing their existing signals to those enhanced with Enigma data.
They considered a business “open” if transactions were present both in the most recent three months and the past 12 months. “Closed” meant a business showed activity in the past 12 months but also went more than three months with no reported transactions.
Results of the pilot evaluation gave the team confidence in the potential of Enigma’s data to increase accuracy across their full dataset.
In a sample study of businesses the team had marked closed, Enigma data found:
- 60% were verified closed
- 40% should have been flagged as open
On an ongoing basis, Enigma data would give the team more accurate signals to verify business status for 18M businesses nationally, without manual research