Automating data entry


2023-24

At its core, Magical was a text expansion app helping users crush repetitive writing tasks. To increase our footprint, we knew we needed to automate data entry too.

We focused on developing a customizable solution for smaller teams that couldn’t afford bespoke, API solutions.


My role

I led all product design work from ideation to shipping.

I collaborated very closely with two PMs and six engineers. This project was ongoing, lasting around 7 months.  


What is Magical?

Magical is a browser extension that helps users automate mundane tasks.

Through text expansion and data entry, 700k+ users saved 7+ hrs per week (on average). 

Most common user types: sales people, recruiters, and customer support. 



How do most of our users move data across the web?

The answer = copy and paste. 

Copy and pasting pieces of data across sources (ex: LinkedIn) and pasting into destinations (ex: Sheets) is a foundational, daily task for recruiters, sales people, and customer support.  

It takes a lot of their time. And they’d rather be doing other things. 











Existing experience


Before I joined the team, Magical launched a first- pass for data automation.

The main problems were:

1. Users had to context switch between sources and destinations.

2. Site labeling was complicated.

3. The dropdown suggestion ordering was horrible.

4. When we did autodisplay, we got in the way









Our users hated this feature.






“I would rather copy and paste than have to use this product. This is more monotonous.”
- Kyle, Customer support







To not tank our ratings, we pushed a quick fix to the site info editor.





Quick redesign focused on keeping users in “edit mode”





Although refinements helped, we knew this feature needed to fundamentally change.

We naively assumed that if we made site labeling clearer, then users would successfully use the dropdown. 

They didn’t because Magical wasn’t integrating into their workflows. 









Research



Business goals

Increase engagement
Only ~15% of our user base had successfully filled a piece of data.

Train a suggestion engine
The previous engine didn’t learn from user behavior. Because of this, it would often suggest company name for a first name field. Not great. 



User research


(25 users, 45min each)


My PMs and I wanted to hear first-hand what users did and didn’t like about our current product. We walked through their current workflows, including and excluding Magical.

We focused on their day-to-day workflows.



What users wanted:





What users didn’t want:













Testing concepts

How should we label site info?









How should we design the dropdown?






No matter how good our UI was, our UX was contingent on high quality suggestions.

We spent months testing / re-building our suggestion model.  










Final designs







The dropdown 






Activating users while they work









Activating before while they work







Results

+250% in activation*



This feature was our most successful activation play to date!

*anytime a user accepts a suggestion (ex: clicking on a dropdown suggestion)


“Too intrusive” drops from #1 uninstall reason

When users delete their Magical account, we ask why they’re leaving. Before this feature, “too intrusive” was our #1 reason for uninstall. 

After this launch, it dropped out of the top three.


Buyers were willing to pay more for Magical


Potential accounts signaled they were willing to pay “2x more” for Magical with this feature set.