Is there such thing as a perfect decision? At FICO World 2014, FICO CEO Will Lansing asked the audience of 950 attendees that very question. He conducted a thought experiment asking, “What might happen if constraints around data went away? If we could get as close as possible to perfect information about every customer?”
The answer is that we could make more perfect decisions and truthfully we’re not that far away. For this to happen, Lansing suggested that you would need access to Big Data, an integration platform, an ETL, analytics and a robust deployment environment.
This idea of the perfect decision opens up all kinds of questions. For example, does a perfect decision have to be real-time? According to Lansing it does, but you have to recognize the limits of that. During a credit card transaction, you’ve got 10-15 milliseconds to run your analytics. There’s only so much you can jam into that time. He pointed out that FICO’s fraud management technology runs about 15,000 analyses in that time, but even that isn’t everything.
He pointed out that when a customer comes to a retail site today, that the retailer has a data challenge. The retailer may have tens of thousands of customers, hundreds of thousands of SKUs, and perhaps 12 slots on the website to load for the customer – what is the retailer going to offer that customer?
In order to do this in real time, the retailer will need to run propensity models on millions of their customers, hundreds of thousands of SKUs, depending on objective function and populate the 12 slots based on highest propensities and the retailer’s goals. The retailer may focus on maximum conversion. But at other times they may want to increase lifetime value, so they load items that take the customer on a path toward higher purchases in the future.
Lansing described how to get to real-time -- you must do a lot of the calculations upfront. Then in those15 milliseconds, when you know the customer is looking to buy something, the retailer can use rules to do an inventory check.
This hybrid approach between pre-loaded calculations and real-time analytics, is just one example of what it takes to make the perfect decision. But technical considerations are not the whole story.
The Three C’s
“We’ve been in the decision modeling business for nearly 60 years now,” Lansing said. “We get that two of the most important pieces of the perfect decision are not about the speed of your technology, or even the computational power of your analytics.”
Lansing describes other, “softer” but vital considerations as the three Cs: Context, compliance and collaboration.
- Context asks the question: When, where, why you’re making a decision and what you want the outcome to be?
- Compliance is the understanding that data-driven decisions about people have consequences, and compliance is part of that.
- Collaboration looks at how you can involve all of the parties with a shared interest and that includes customers.