Banking apps tend to be uninspiring. Nowhere is that more apparent than in their presentation of transaction data. Transaction details are usually a dry collection of technical data—often full of banking jargon and elusive abbreviations.
At Numbrs, our goal is to make banking more approachable and more insightful. In search for impactful ideas, I initiated a project to transform our original transaction details into a more meaningful and modular feature.
Due to my NDA, I can only share selected details of the project—no research, strategy or business background.
The original transaction details screen was limited to the essential data we received from the banks, such as counterparty, amount and subject (or reference in the UK). Very similar to your run-of-the-mill banking app.
At the same time, we saw that people were less interested in these descriptive details and more in seeing the bigger picture.
Recognising a transaction is difficult—especially weeks or months after it happened. Banks use abbreviations and ambiguous conventions that are not human readable. With our data science team—hat tip to Antonio and Bartosz—, we identified that there’s a lot of partial information here and there that can be used to determine the type of a transaction (eg. card payment, direct debit, etc.) and to provide other meaningful metadata.
Apart from clarifying the data, I found that one of the most useful ways to make transaction data more understandable is to focus on habits and series rather than single purchases. This makes it easier for our users to grasp their own spending and saving patterns. Instead of looking at a single purchase at, for example, Tesco, you will see your full purchase history, your average spending, how many times you shop there, etc.
The new designs organises all information on a single screen:
My goals to guide the design process were:
Some examples for the transaction description variations:
Apart from systematic differences, I also designed for the various usage patterns we saw analysing behavioural data:
To properly organise and analyse income and spending, transactions need to be sorted in the right categories first. Even though our algorithms will try to classify each transaction automatically, in some cases we rely on the user to manually pick a category.
Realising that the classifier will mostly have some bets—matches with lower weights—, I created a concept to show these matches as suggestions. After validating the idea, I redesigned the category picker adding suggested categories. The project made manual categorisation a lot faster, users selecting from the suggestions even more often than we have anticipated.