Process: Research, Interviews with Subject Matter Experts, Competitive Analysis, Mockups/Prototypes
Solution: Multi-page guided process allowing easy configuration for 80+ fields to quickly import 10k records without extensive
Result: Greatly reduced failure rate, decreased time needed to onboard
For a long time, Import Assistant was the most difficult and complex thing I worked on at Chase but last year, the work I did on the Number surpassed this. What I’m saying, this had a lot of moving pieces.
Problem: There was a need to allow customer to upload lists of records (many thousands) into our software / system to enable easier management of account and significantly reduce technical issues on the customer’s side. Self-service is frequently a goal in the work we do. There was an existing process for the customer to use but it was extremely difficult and required the customer to have a tailored database output the precisely matched our strict requirements. This was prone to failure.
Detailed Requirements: We needed a method to import a file, to map the fields of the file to fields of our software, a way to use existing data or choose from existing data, a way to save progress, and a different way to save the finished map.
Process: Everything starts with research. I needed to better understand the existing process, the people using the process, and what data we were even working with. To make it more difficult this all began in the first couple weeks at a new job. I spoke with a lot of internal people from support to subject matter experts to people who previously did this work as customers are difficult to interview / interact with in this environment to better understand what goal we were working towards. I spent days scouring for examples from people who had solved this problem in past. I knew about Microsoft Mail Merge which did something similar but not to the extent we were aiming for and I found other software which could take .csv or .xlsx files and import them to systems but nothing I found was close to meeting all our requirements.
When I had enough information I began writing out use cases and sketching rough information architecture layouts to test with. Most of this testing was solo but I also engaged the subject matter expert when any progress was achieved. Below you will find sketches as I worked to organize and build the architecture to support the information needed. One of the first successes I had was being able to combine two of the columns into one, reducing the complexity of the page markedly:
Once I was confident in the sketches, it was time to hit Axure. The module would require multiple page-states to address various configurations and entry paths into it. I met with the development team every morning for stand ups and twice weekly for refinement sessions to ensure everything was progressing without issue.
I next worked on the interaction design or how do we manage the immense amount of information that can be flowing through the 80+ fields that the process requires. After establishing the types of data each column would hold and the way the user would interact with it we were faced with one the impossible questions. Establishing format of the data or how do you know if something is MM/DD/YYYY or some other variation of it: YY/MM/DD, MM/DD/YY… Where there was hope we could do this programmatically, the team had to accept this was impossible and move these impossible items to earlier in the process. There were a couple of these items but date is the most recognizable to people. All finished work would also need to adhere to WCAG 2.0 accessibility standards.
For page copy I worked with our Technical / UX writers. Axure screenshots below:
And it’s live!
Follow up: I returned to this project months later to add additional functionality in managing completed maps and searching for templates to use. Feedback can be difficult when one works so deeply inside enterprise software but of the people I know who use this, none have ever complained to me about this. There is overlap here with people who use the work produced from the Number work.