Everything Stays Different

Figure 1: Landing page variants for HASPA personal loans

Every person is different, and everyone has different needs — that much isn't new. Personalization allows companies to address their customers individually, based on those needs.

Let's look at an example: Susanne is 30 years old and lives in downtown Mannheim. She wants to buy a car so she can visit her boyfriend in Berlin more often. When she visits her bank's website, she sees a different homepage than her neighbor Andreas, who has long dreamed of a house in the countryside with his family.

Based on parameters such as location, interests, and the device being used, Susanne's bank can serve the website variant that best matches her needs.

Conventional personalization solutions rely on A/B tests. They test different variants against each other and select a preferred variant once statistically significant differences emerge. Once that distribution is settled, the learning process is complete.

Machine Learning

Acceleraid's personalization solution takes a different approach. Machine learning uses all available data parameters and factors in context, enabling far more precise personalization. What's more, the machine learning algorithm detects trends and seasonal effects. It can even be configured to "forget" what it has learned after a defined period and respond to current developments instead. This removes any risk of a single context or trend permanently overriding all future developments.

Why Does This Distinction Matter?

Because needs change — not just from person to person, but from situation to situation. Needs aren't a static basis for categorization; they're dynamic and shift unpredictably.

That's why good personalization isn't timeless and doesn't simply stop after a single learning cycle. Susanne, for instance, might decide right now to move to Berlin to be with her boyfriend. She already works remotely, after all. She'll soon no longer need a car of her own. Instead, she'll probably want to renovate parts of her boyfriend's apartment. Andreas, meanwhile, might feel unsettled by the ongoing pandemic. Perhaps he decides right now that it would be wiser to postpone the search for a plot of land and keep renting a little longer.

Example: Hamburger Sparkasse

We're currently seeing exactly these kinds of shifts up close and in detail with our major banking clients. Together with Hamburger Sparkasse, we ran a personal loan campaign between October 2020 and March 2021 featuring six variants. The messaging ranged from fulfilling dreams — "Make your dreams come true now" — to messaging that foregrounds more fundamental needs: "Live simply." Each user is shown the message that best matches their current needs.Looking at which variants were served most often, we can observe how the underlying motivations for personal loans shifted within a short period of time — from the desire to fulfill big dreams to the need for future security.Figure 2: Delivery of landing page variants for HASPA over timeThe analysis shows that the most successful messages shift in a rhythm that's impossible for humans to predict. In October, the "Live simply" message performed better. In November, more people responded to the dream-fulfillment message. In December and January, "Live simply" dominated again, while in February and March, the dream-fulfillment message performed best. Why needs shift so unpredictably is something even machine learning can't explain — but it can detect that they're changing and respond to them in real time. That way, every customer receives the right offer at any given moment.