CBP Exploring Use of Analytics to Speed Up Rulings Process

SAN ANTONIO -- CBP is looking at using "prescriptive analytics" as a way of making the rulings process faster, said Brenda Smith, CBP executive assistant commissioner-trade, at the National Customs Brokers & Forwarders Association of America's annual conference on April 17. Such analytics involve a review of the data that "puts out a couple of options for action," she said. While "we don't want a machine doing classification," CBP would like to use technology to improve efficiency on rulings, she said.

One reason that some rulings take too long at CBP is that many involve lengthy reviews of past agency rulings, she said. "If you apply analytics to the vast number of rulings that have already been made -- hundreds of thousands of rulings -- as well as the information we have on tariff classification, there's a pretty good bet that you could come up with, using analytics, a couple of options of what the right classification is."

There are multiple data "projects under an innovation lab," she said. "If you look at the massive amounts of data that CBP collects from you all, we, if we have the right tools, could do very detailed summarization of the trends and patterns represented by that data," she said. That would be helpful for looking at trade flows, but also for seeing "who doesn't fit within the general pattern of trade," she said. "Who is selling things at a very low value? Who is bringing in shiploads of things with a somewhat odd quantity?" Predictive analytics will also show "what is likely to happen next," she said. For example, an antidumping order may result in goods starting to come from a country where there were no previous production facilities. "That's something we want to know about and probably ask a few questions," she said.

One project Smith mentioned is an effort with Johns Hopkins University researchers called Socrates (see 1704200009). From the set of data CBP gave the researchers, "we found over 20,000 entries that indicated that there was some form of identity theft underway," she said. CBP also partnered with "a private-sector consortium that helped us combine open source business data along with some interesting technology," and it found that 10 percent of the entities of the dataset were at risk of going out of business and 2 percent were "associated with some sort of criminal activity," she said.