Everything remains different

Figure 1: Landing page variants of HASPA personal loans

Everyone is different and everyone has different needs… so far, nothing new. Personalization enables companies to address their customers individually according to their 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 goes to her bank's website, she finds a different home page than her neighbor Andreas, who has long dreamed of a house in the countryside with his family.

Based on various parameters such as her location, her interests, and the device she uses, Susanne's bank can deliver the variant of its website that best fits Susanne's needs.

Conventional personalization solutions rely on A/B tests. They test different variants against each other and, when statistically significant differences are found, select a preferred variant. As soon as this distribution is established, the learning process is complete.

Machine Learning

Acceleraid's personalization solution takes a different approach. Machine learning uses all available data parameters and takes the context into account. This makes much more precise personalization possible. In addition, the machine learning algorithm detects trends and seasonal effects. It can be configured to “forget” what it has learned after a previously defined period and to react to current developments. There is therefore no danger that one context or one trend overshadows all further developments.

Why Is the Distinction Important?

Because needs change. Not only from person to person, but also from situation to situation. Needs are not statically categorizable bases for decisions; they are dynamic and fluctuate unpredictably.

Good personalization is therefore not timeless and does not simply stop after one learning cycle. Perhaps Susanne decides at this very moment to move to Berlin to be with her boyfriend. She works from home anyway. She will soon no longer need a car of her own. Instead, she probably wants to carry out some renovation work in her boyfriend's apartment. Perhaps Andreas is unsettled by the ongoing pandemic situation. Perhaps he decides right now that it would be more sensible to postpone the search for a building plot and to keep renting a little longer.

Example: Hamburger Sparkasse

Right now we can observe exactly these fluctuations firsthand and in detail with our large customers in the banking sector. With Hamburger Sparkasse, we ran a campaign for personal loans with six variants between October 2020 and March 2021. The motifs range from fulfilling dreams: “Make your dreams come true now”, to motifs that put elementary needs in the foreground: “Just live.” Each user is shown the motif that individually best fits their current needs.Based on the preferentially delivered variants, we can observe how the motivations for personal loans have changed within a short time. From the wish to fulfill big dreams to the need to create security for the future.Figure 2: Delivery of the landing page variants for HASPA over timeThe analysis shows that the most successful motifs change in a rhythm that is unpredictable for humans. In October, the motif “Just live” was more successful. In November, more people were interested in the dream-fulfillment motif. In December and January, the motif “Just live” dominated, while in February and March the dream-fulfillment motif was the most successful. Why needs change so unpredictably cannot be explained by machine learning either, but it can recognize that they are changing and respond to them in real time. This way, every customer receives a suitable offer at any time.