4.4.2 Beyond sampling

• Sampling was a product of a time when obtaining and processing data was expensive, both in terms of time and money
• Modern computer technology means that the cost of acquiring, storing and processing data has fallen dramatically
• Rather than sampling a small subset of the data, it is now possible and affordable to acquire data on the off-chance it will be useful; and then either examine extremely large samples, or indeed process all the data
• Big data analyses huge pools of data to find new connections that would have been missed by randomly sampling a tiny fraction of all data
• Big data


  Analysing huge pools of data to find new connections that would have been missed by randomly sampling a tiny fraction of all data
• Fraud accounts for a tiny fraction of all transactions, so it is highly unlikely a sample of transactions taken from the many millions of transactions made every day would include a single case of fraud
• By examining all transactions for unusual patterns of activity, potentially fraudulent transactions can be identified for further investigation
• Fraud detection


  • The financial company Xoom specialises in remittances: money sent by immigrant workers to their home countries
  • This is a business which has traditionally been vulnerable to fraud and money laundering
  • Xoom's automated systems scrutinise every one of the company's transactions
  • In 2011, they identified a string of payments originating in New Jersey made on Discover credit cards
  • None of the transactions were themselves unusual, but the pattern created by the times of the transactions, the sums transferred and their recipients were sufficient to arouse suspicion
  • Xoom's fraud detection team were alerted and found the transactions involved criminal activity; the payments were blocked, no customers lost money and the criminals were identified