Statement's financial platform relied heavily on both automated and manual data enhancement. When our data team developed a new algorithmic model that allowed for clustering similar transactions to enhance them collectively, our mission as the product team was to create a flow that would not just expose this option to users, but actively encourage them to pursue this more efficient action
Why was this so important to us?
The value of this capability may seem trivial, but in our case, it was crucial. Accurate transaction tagging directly supported the robustness of other modules in the app, such as reconciliation, forecasting, and more.
The algorithm
The algorithm grouped transactions with similar bank-provided attributes and selected a small sample from each group. Users could approve these samples to update the entire group, but partial approvals posed a challenge—updating only part of a group wasn't feasible. To maximize accuracy, we allowed changes to the approved sampled transactions while leaving the rest untouched. This led to three sub-flows:
Approving all samples → updated the full group
Approving some → updated only those and the original transaction
Approving none → updated only the original transaction.
Flow’s starting point
For all the solutions we explored, the flow began the same way: when a user views their transactions and chose to change the category of a single transaction, this action triggers the option to apply changes to other, similar transactions the algorithm had identified.
Defining and Designing Process
While this wasn’t the most complex feature in terms of screens or components, it did come with a few UX challenges, and a few things that we had to keep in mind:
Straying away from original plan
Users had an initial intent that should be respected, otherwise they can get overwhelmed, confused, and irritated.
Note to self: always boldly allow for the narrow change of only the original transaction.
Dynamic and non-universal consequences
The exact scope of the action is dynamic, not exactly based on any universal or pre-known logic, and therefore can be confusing.
Note to self: the reflection of the exact result should be presented clearly, and should also change according to current selection at all times.
Extensive impact
There is a certain amount of details that are important for the purpose of making a correct decision, which have to be fully and clearly presented.
Note to self: communicate the exact results of the action, and do so in a positive and encouraging, yet very precise manner.
Important details
There is a certain amount of details that are needed in order to make a correct decision, which have to be fully and clearly presented.
Note to self: Include all relevant details and show them in a clear structure, preferably one that’s familiar to the user from the platform.
More “now” work
Even though this action saves time and work in the long term, it increases immediate manual work.
Note to self: use visual cues and well-put copy to convey and emphasize the value of this bulk action.
Our Final Three Options
We explored three different design approaches for this flow, weighing the pros and cons of each. Here’s a look at all three and the reasoning behind our final choice.
Solution 1: The One-Step Approach
In this solution, we wanted to capitalize on the momentum of the user’s initial action. To keep users engaged and reduce the risk of them abandoning the flow before completing the bulk update, we decided to present everything in a single popup.
We knew this could create a high cognitive load, so we took steps to ease it:
Clear visual separation between the two options—updating just the original transaction or applying the change to the larger group.
Bold, distinct call-to-action buttons in each section to guide users toward their desired action.
Despite these efforts, the interface still felt tight and somewhat overwhelming, which led us to explore alternative solutions that might provide a smoother, more intuitive experience.
Solution 2: The Three-Step Approach
In this approach, we aimed to reduce cognitive load by breaking the flow into three clear steps:
The user updates the original transaction as intended.
They receive a validation to this change, accompanied by an engaging suggestion to proceed to updating the sampled transactions and the larger group.
If they choose to continue, they review and confirm the bulk update.
This structure made the process feel less overwhelming, as users could focus on one decision at a time. However, we were concerned that users might not feel motivated enough to proceed beyond the first step, potentially missing the opportunity to apply the bulk update.
To improve our chances, we incorporated:
Engaging visuals to create a sense of automation and ease.
Clear, compelling copy to highlight the value of updating multiple transactions at once.
However, we felt this additional step was still at risk of being skipped altogether, leading us to explore alternative solutions that kept users more naturally engaged.
Solution 3: The Two-Step Approach (Selected Solution)
This solution struck a balance between the first two approaches, aiming to minimize cognitive load without compromising user engagement. It offered a middle ground between the efficiency of the one-step flow and the clarity of the three-step process.
After the user updated the original transaction, they were immediately shown a success message that seamlessly introduced the option to apply the change to the sampled transactions and the larger group. This allowed users to:
Easily complete their initial task without distraction.
Be prompted right away with the broader action, keeping them in the flow without adding friction.
The key advantage here was that users didn’t have to pause or overthink their next step—the option to execute the bulk update was presented naturally, requiring minimal effort to proceed. This approach provided just enough space to keep the design clean while maintaining a smooth, intuitive flow, which is why we ultimately chose it.
Reflections & What Possible Next Steps
In an ideal world, we would’ve done some A/B testing to compare solutions and see how users responded. But in the fast-paced environment of a small, agile startup, this wasn’t our typical process. Although I wasn’t with the company when the feature launched, it would have been valuable to track things like: Completion rates for bulk updates. How long users spend on the popup—are they lingering or deciding quickly? Whether there’s a learning curve, with users getting faster as they become familiar with the flow. These insights could’ve helped fine-tune the experience based on real-world usage.