July 4, 2019 by Jason Contant
Artificial intelligence and machine learning are changing the landscape for pre-fill data in Canada — literally.
“Through artificial intelligence and deep learning techniques… we’re able to train computers to recognize that this is a two-storey structure, not a one-storey structure or a bungalow or a ranch bungalow,” Greg McCutcheon, president of Opta Information Intelligence, said by way of example. “We’ve been able to train computers to recognize the type of construction and features about that home by gathering information through imagery.
“That’s a modern update that we didn’t have two or three years ago.”
McCutcheon spoke to Canadian Underwriter last week about the status of using pre-fill data for home insurance application forms in Canada. A standard form can have roughly 40 or more questions, he observed. The Canadian property and casualty insurance industry is currently focused on trying to determine if they really need to ask all of those questions, or whether pre-filled information can be used.
And so where are we a with pre-fill data today? More specifically, how accurate is the data?
McCutcheon said it’s difficult to say exactly. Sometimes it can vary by geography. Generally speaking, the accuracy of the data has improved over the years.
“When we started the validation process with the data, sometimes two or three fields had to be changed or tweaked,” McCutcheon said. “Often what we are seeing now is, particularly in urban centres, the information is 100% accurate and there is nothing to change.”
In rural locations, machine learning techniques are used “in a bigger way” to capture information from satellite imagery. “We’ve really had to modify and zone in on how are we able to harvest data in those areas to make sure the data is as accurate as urban centres,” he said.
Almost a decade ago, Opta launched its property insurance valuation service iClarify, which pre-fills data in approximately 90% of all new business quotes in Canada. Data is pre-filled into all the major broker management systems so that brokers already receive 12 pre-filled data fields of construction features; brokers can quickly validate these features to get the replacement cost value of a property.
Opta does about 30,000 individual new business quotes through iClarify on a daily basis. The data is updated regularly, based in part on consumer response. “If we have pre-fill data that says the basement isn’t finished, for example, but a consumer tells us they have a finished basement, we accept that change and we instantaneously correct the data,” McCutcheon said. “On behalf of the industry, we are updating this data set daily based on these conversations, which makes the whole process for the entire industry better.”
Opta employs full-time staff solely to harvest information, correct it, and get updates through the insurance quote process to drive accuracy. “That information is updated through the evolution of the nature of the dialogue with our customer base,” he said. Accuracy also comes from employing a recently-reunified inspector team, Opta Precise Services, to capture information while they conduct property inspections.
When procuring a dataset from a third party, Opta checks that dataset against “what we know when we inspected the property and we measured it off,” McCutcheon said. “That’s a standard on how we update our datasets.”
Ultimately, however, “people need to verify any information source that’s being pre-filled, as conditions to properties do change,” McCutcheon added.