Build to learn
One of the key things that differentiates a design technologist from a production developer is that our core metric is shipping a great experience. Sometimes everything flows smoothly from design, to testing, to building and shipping the design vision, other times most of the code we've written ends up on the cutting room floor. Prototyping to learn we shouldn't move forward in a certain direction frees us to focus on more impactful work.
End-to-end QuickBooks Prototype #
Early on, our team collaborated on an end-to-end prototype for QuickBooks that spanned from marketing to initial setup, importing data, invoicing, tracking expenses, and much more. The entire team contributed to this mega-prototype at various times. A key challenges for us was collaborating on a shared project when we'd mostly been working fast, dirty, and independently in the past. Suddenly we had to consider other coding styles and state management strategies, plus just making sure our code could be easily digested by someone else.
One List #
Many of the prototypes we built so people could test experiences with real dynamic data without concerns about data protection and privacy. Building outside of the production platform lets us deliver much more quickly and with less project complexity.
The goal of the One List project was to address the fact that QuickBooks' biggest competitor isn't other accounting software options, it's ...drumroll... Excel. The intent was to give customers more flexibility in how they parse their data and create the potential for new insights by allowing granular filtering and searching. Additionally, a frequent pain point for new customers was not being able to do the same things in QuickBooks that they could do in Excel.
One area where we paid particular attention was the date picker, with the ability to view data for a week, day, or month and jump forward or back a single unit without reopening the full date picker.
Learnings #
User testing with this prototype revealed a few key findings:
- Navigating through the data by weeks was a good match for existing mental models.
- Many of the accounting terms weren't meaningful to people coming from Excel (Bank Transactions, etc.).
- Showing the filters in the sidebar created the expectation that they were applied independently from the date window + search query. In reality all filters were applied simultaneously.