"Interest rates: Nubank studies how to reduce or even end with the revolving rates"
We're always designing products.
But what about deleting some of them?
Credit card's interest rates in Brazil are very high!
That's why one of the most important projects for Nubank in 2020 was to reduce the average interest rates that we were offering to our users. At that time, Nubank ranked 41st (Nu Financeira) on the Central Bank's list of "the highest rates":
Workshops
To assess our opportunities, my team and I started to run some workshops to understand our options. To begin, we knew we had 3 different possibilities:
(01) New product proposals
(02) Existing products optmizations
(03) Additional financing options
But, one of the ideas that emerged from the workshops was completely different from those 3 topics: what if we just "delete" the product with the highest interest rates?
At first, it didn't seem quite smart to just delete one of our products. Because, like the old expression would say: "you don't solve the problem just by ignoring that it exists" (I just made that up). But, it wasn't about ignoring the product, it was actually deleting the product. And what if that was exactly what was holding us back?
Assessing the problem
After accepting the idea of "deleting" the product, my PM and I started another product assessment document. Our idea was to write about:
(01) The context
(02) Problem hypothesis: what are planning to deliver? How does it connect to our segment and unmet needs?
(03) What we already know
(04) Questions that we need to answer – divided by cluster (marketing, data, product) so we could find owners to every question.
Defining experiments to validate the hypothesis
After writing down everything we knew and wanted to discover, we started our Discovery Agenda, and the first step was to understand the experiments that we would do to validate our hypothesis.
To do that, I invited everyone on the team to a workshop on Miro where I put every single question on a post-it and the team had 15 minutes to think and write a possible experiment solution that would answer that question.
Question example: "Which customer segment is most sensitive to high interest rates?"
Experiment example: "We could run a quantitative research using the Sean Ellis method to validate our segments first hypothesis".
And after this brainstorm, we discussed idea per idea and I gave them 5 minutes to vote on the most important ones.
We discussed again the voted experiments and decided to put them into a "Effort vs Impact" matrix.
The "removing the feature" idea was in the low-effort, high-impact quadrant. So let's do it.
I started by writing the experiment definition (with a simple framework that I always use):
(01) Hypothesis
(02) Objective of the experiment – what are we aiming to discover? Why are we doing this? Which company OKRs will be affected by the features proposed in the experiment?
(03) How to do it – defining control, variable and treatments.
(04) Sample target – which segments? how many?
(05) Possible risks
(06) How do we measure success?
(07) Duration of the experiment
(08) Results (to be completed after the experiment)
And that's how the "Non-revolving" experiment started. We were calling it that way because the product we wanted to end with is called "Revolving" (when you don't pay your credit card bill in full, but does at least a minimum payment and what's left is added to your next bill with a 14,75% interest rate!).
The experiment
Ok. We had our hypothesis and our experiment to validate it and in Aug-15, we rolled out a non-revolving test for 150k customers where the goal was to assess viability from an economic and user satisfaction standpoint of offering a credit card without the revolving option, but when the customer needed to revolve, we would use our "instalments plan" product in place of the one that we deleted.
Note that, for scrappy test reasons, our experiment were designed only for new clients – which wouldn't exactly help us to decrease the average interest rates, but would give us a hint about it.
First results and business impact
After a month, we had a 38% reduction in the interest rates (9.43% last estimate for non-revolving vs 13.23% in our average portfolio).
Next steps
I was also responsible for a qualitative research with our users to understand what was their feeling about the product we were offering, and alignment with the operations team to understand possible backlashes (which led us to find a very important corner case for the iteration of this non-product).