When I joined Wave as a product designer on the Accounting team, I inherited a quiet but critical problem: users weren’t engaging with a core feature—transaction categorization—as much in the new version of the app.
Categorization isn’t just a nice-to-have. It’s foundational to bookkeeping. Every transaction needs to be assigned to the correct account in the Chart of Accounts. If this workflow breaks down, so does the integrity of a business’s financial reports.
With a platform migration on the horizon, we needed to understand why engagement had dropped, and how to fix it.
Company
Wave Financial
Summary
Cut clicks by 40% and brought clarity back to a core accounting task. Rebuilt 9 years later with Lovable.
Year
2016, 2025
Process
Zooming in: Understanding the workflow firsthand
I started by walking a mile in our users’ shoes.
Using my own bank data, I categorized transactions in both the legacy and redesigned versions of the product. The difference was immediate: the legacy experience was fast, compact, and intuitive. It used a dense table layout where rows expanded inline—minimal clicks, minimal thinking.
A screenshot of the legacy version to give you an idea of the difference in the two version’s experience
The new version, however, introduced friction. Interactions were split across panels and dropdowns. What used to be a smooth, focused task now felt fragmented.
Seeing this contrast firsthand gave me an early insight: this wasn’t just a visual redesign issue—it was a workflow regression.
Finding the friction
To dig deeper, I turned to FullStory to observe real user behaviour at scale. I focused on how users interacted with the categorization UI, especially the two dropdowns that were required to complete the task.
What I saw confirmed the friction I felt: users would start categorizing but often abandon the process midway—many didn’t make it past the first dropdown.
This was a red flag. Something in the new flow was increasing cognitive load and breaking user momentum.
The two dropdowns required to categorize a transaction. See full design of the page below.
Prototyping solutions: from surface tweaks to structural change
We started with a light-touch experiment.
1. Highlighting incompleteness
I changed the “Uncategorized” label to red to draw attention to transactions that needed review. It helped users notice what needed action—but not how to act. The gap between noticing and completing the task remained too wide.
2. Rethinking the flow
Next, I proposed a deeper change: consolidate the two dropdowns into one, and bring the categorization interaction inline within the transaction list. No more bouncing between panels.
The goal was to:
Minimize context switching
Reduce clicks and cognitive overhead
Help users complete the task in one focused motion
I tested both prototypes on UserTesting.com, giving participants common bookkeeping tasks to complete, comparing current workflow with the proposed design. The results were clear: the new inline flow was faster, more intuitive, and better aligned with user expectations. Internal stakeholders such as support and finance teammates agreed.
Shipping the fix: simplicity by default, detail on demand
With alignment across design, product, and leadership, we shipped a redesigned Transactions page that brought focus and flow back to categorization.
✅ The compact list view became the default
✅ A simplified single dropdown replaced the two-step categorization
✅ A side panel remained accessible for more detailed edits, but was no longer required for basic categorization
The changes made the experience feel more like the legacy version in terms of speed, but more modern and scalable in implementation.
Outcome
The result: a return to confidence
This redesign helped restore user engagement and trust in a foundational workflow. It also gave the team confidence that we could evolve the product without losing what made it work in the first place.
📉 Friction reduced by 40% — categorization went from 5 clicks to 3
📈 Increased completion rates for categorization
What I took with me
Design is a system of tradeoffs. Moving fast can sometimes mean accidentally breaking the wrong things. Recovery means listening, testing, and not being afraid to revisit decisions.
Usability testing isn’t just validation—it’s insurance. It’s what gave us the confidence to challenge a recently shipped design, and the evidence we needed to bring leadership along.
This work was one of my favourite examples of solving a real problem by blending behavioural insights, rapid testing, and interface clarity—ensuring that what’s good for users is also good for the product.
9 years later: AI redesign
Just for fun (and to see how far I’ve come), I revisited a page I designed nine years ago using Lovable and AI-powered categorization. The redesigned experience cuts the workflow from three clicks down to just one—or even zero—for medium to high confidence AI suggestions.
Read my LinkenIn post → about my process and learnings so far.
