Auto Categorize Expenses

Auto Categorize Opt in screen

 
 
 

What does Auto-Categorize do?

Auto-Categorize is a cross-platform feature that uses Machine Learning and AI to auto-categorize expenses for QuickBooks customers. When our engines reached 80% accuracy, I proposed this feature to my team.

Benefits & Risks

Benefits:

  • Expectation - AI and Gen AI's widespread use has raised user expectations significantly.

  • Saves time - Categorizing transactions stands out as the most time-consuming activity for users

Risks:

  • AI Mistakes - Many small businesses have a very low tolerance for AI mistakes.

  • Brand damage - If AI makes errors in managing their books, it can erode users' trust.

I established two principles to mitigate the risks:

  • Educate users - Before deciding, users should know the features' costs and benefits.

  • Let users review the result - Users should be able to view and edit the outcome.

My hypothesis: If we offer customers the option to have QB Auto Categorize transactions it’s confident about, customers will opt-in.

Leap of Faith Assumptions:

  • Customers understand and are invested in training QuickBooks for the long-term benefit of more transactions being categorized more accurately over time.

  • Our categorizations are accurate enough to make it worthwhile for customers to review and fix what's been automated (vs. manually categorizing the current way).

  • The customer feels like they have trust and control over the automation.

  • The customer feels like this experience will offer benefits by saving them work.

Desktop Version

I started by designing the desktop version because:

  1. Our Desktop engineering team was available to pick up a project

  2. We had more research and understanding about desktop-only/mostly customers

  3. The number of desktop-only/mostly customers was significantly higher than that of mobile users

Mobile version

For designing mobile, we first followed the same logic as the web, but we ran concept testing with customers to ensure that it resonated with our mobile customers.

Research Method & Recruiting Criteria

  • 6 US Small Business owners who manage their accounting on their own

  • Categories transactions on the QuickBooks mobile app at least once a month

  • Has ≥30 transactions a month - Ranged from 30-500 transactions a month

Soon after testing the initial concept with real QuickBooks Mobile customers, we learned that users' mentality is very different when using web vs. mobile. The same person spends more time reading, reviewing, and thinking through activities while using a desktop than a phone, mainly for accomplishing specific tasks with a meager attention span.

After testing the first concept with six mobile customers, we iterated on the prototype after some themes emerged from our initial interviews and tested again with six more customers.

 

Research Findings

 

Which concept to pick?

Version 2 - Preview list first

Pros:

  • Higher chance of intriguing customers in one click

  • Show customers what will happen to their actual books

  • Higher adoption

Cons: 

  • Customers may not absorb the feature and its functionalities completely 

  • They may miss the connection between the list and opt-in 

Version 1 - Explanation first

Pros:

  • Set up the right expectations for customers about Auto-add

Cons: 

  • Customers may not read through

  • The information is abstract - it is hard to imagine how it impacts them personally.

  • Losing not tech savvy or anti-automation customers

  • Fewer customers at the top of the funnel

Although the second iteration had more successful outcomes from concept testing with QuickBooks Mobile customers, they were not fully aligned with my design principles. The main concern was that users would opt in without fully understanding the feature's benefits and risks.

I scheduled a work session with my team to review the pros and cons of both concepts and improve the opt-in page's information density. I also ran concept testing with twelve people on Usertesting.com to validate the third version, which was very successful.

Final Figma File

Web & Mobile Version - Comparison View