Past work on for example megastudies finds that experts (scholars) find it really difficult to predict which nudges will be effective. One possible reason is that experts are often really different from the population they try to target with their nudges. We (Anna Paley and I) therefore develop a method to let the crowd generate potential nudges, have another group from the crowd evaluate these nudges, and then implement it in a field test.
We test our method in the domain of consumer oversubscription. Together with Nibud, the Dutch budgeting institute, we find that people have many more subscriptions than they think they have. We then document that if they become aware of this, they would also like to cancel some more of their subscriptions. However, this is normally difficult to achieve because it takes time and effort to make a complete list of your subscriptions.
We teamed up with the Rabobank, a large Dutch bank that was about to launch a feature in their banking app that would provide an overview of one's subscriptions. At the launch, customers would receive a pop up message highlighting the new feature. We used our crowdsourcing method to generate alternatives to that message.
We had 200 Dutch people create possible messages. The bank's experts added a few and we as academics added a few. We then had a different group rate how likely they would be to open the new feature if they would receive the given messages. Turned out that the bank's best message ranked only 68th out of the 154; while our own best one was ranked 100th.
We finally field tested 4 of the top 20 messages, and compared it to the bank had intended to use. All crowdsourced messages did better than the one from the bank, and in the end these versions increased the number of users of the feature by over 60,000.
All in all, this is a first but clear indication that crowdsourced messages can be very effective and outperform those made by experts.
We are now working on follow-ups to expand and further test this method.
We test our method in the domain of consumer oversubscription. Together with Nibud, the Dutch budgeting institute, we find that people have many more subscriptions than they think they have. We then document that if they become aware of this, they would also like to cancel some more of their subscriptions. However, this is normally difficult to achieve because it takes time and effort to make a complete list of your subscriptions.
We teamed up with the Rabobank, a large Dutch bank that was about to launch a feature in their banking app that would provide an overview of one's subscriptions. At the launch, customers would receive a pop up message highlighting the new feature. We used our crowdsourcing method to generate alternatives to that message.
We had 200 Dutch people create possible messages. The bank's experts added a few and we as academics added a few. We then had a different group rate how likely they would be to open the new feature if they would receive the given messages. Turned out that the bank's best message ranked only 68th out of the 154; while our own best one was ranked 100th.
We finally field tested 4 of the top 20 messages, and compared it to the bank had intended to use. All crowdsourced messages did better than the one from the bank, and in the end these versions increased the number of users of the feature by over 60,000.
All in all, this is a first but clear indication that crowdsourced messages can be very effective and outperform those made by experts.
We are now working on follow-ups to expand and further test this method.
