Most B2B organizations probably treat their customers as slightly more mature than the recently-born Archie Harrison Mountbatten-Windsor, but Forrester VP and principal analyst Mike Gualtieri suggests they should hold just as much of an interest as “Baby Sussex.”
Speaking at an event focused on digital decision-making hosted by analytics firm FICO in Toronto on Wednesday, Gualtieri said brands that truly want to be more customer-centric should not merely try to create good experiences, but “royal” experiences, offering the kind of personalization that would be fit for a king, queen or other member of a crown-wearing household.
“We know just about everything about them,” Gualtieri pointed out, showing a slide of tabloids and comparing it to the rapid way in which smart companies pore over customer data. “They learn about their individual characteristics, detect customer needs and desires and adapt applications to serve them in real time.”
Of course, doing all that isn’t easy, which is why Forrester last year introduced a category it calls digital decisioning platforms, which combine all the necessary technology to achieve royalty-level customer centricity. In its “New Wave” report released last October, for instance, the company listed FICO alongside IBM, SAS and nine other vendors.
Beyond choosing a platform, however, Gualtieri suggested application and analytics professionals will need to educate themselves, alongside their counterparts in other business functions, about how to make use of artificial intelligence (AI) approaches like machine learning to radically improve the kind of experiences they offer today.
A common example to help make the business case, Gualtieri explained, was a project that looked at a traditional direct mail campaign to one million customers. Assuming a cost of $2 per mailing to market a $220 service, a 1% response rate would mean 10,000 customers. $200,000. A similar campaign based on machine learning, however, might only send to 250,000 customers, where a 3% response rate would yield, 7,500 customers, or $1,150,000
“(Machine learning) can have a multiplicative effect on profits,” Gualtieri told the crowd. The trick is to understand that accurate models may not exist for every question a business might have, and that the algorithm used — he suggested the audience learn vector machines, gradient boosting, random forests and others — might need additional “training data” over time.
A good example looked at the question of whether a particular group of people in a company would play golf, for instance. Gualtieri showed a series of columns and rows that included variables such as the temperature, humidity and wind. He suggested taking 70 per cent of the rows and putting them into the “training bucket” and 30 per cent into a “testing bucket.” Applying various machine learning algorithms would show whether or not they were getting a good result.
“If the application is only 50 per cent accurate, that’s no better than a coin toss,” he said. “Instead of another algorithm, go back to business people and ask if there’s anything missing in the data set.” This could include the day of week, skill level of the players, the condition of the course and so on.
Though analytics professionals are used to code always running as it is written, Gualtieri said they and other business leaders need to realize that models in AI need to monitored and retrained as they degrade over time. They also shouldn’t always feel compelled to do what the algorithm tells them to do, he added.
“Encapsule it in decision rules with overriding business logic,” he said, adding that a data decisioning platform can provide “guardrails” based on common sense.
Bill Waid, FICO’s vice-president and general manager, Decision Management, noted that too many organizations have limited visibility into customer lifecycle ROI and that many of their existing applications were not designed for customer decisioning. Getting to customer centricity doesn’t have to be accomplished through a single overarching project, he added.
“You can spend billions of dollars to transform, but those efforts don’t necessarily come out any better than those who take an iterative approach,” he said.