Helping people start their financial conversation
My-Fi is a financial chat assistant for Canadians on low and middle incomes. Its 19-question setup stood between signing up and asking a first question. I shortened setup, used those answers to suggest where to start, and made the profile editable so people could add context later.
Skipping setup left people with little to work with.
My-Fi asked about income, debt, work and education before people had tried the chat. “Skip for now” ended the questionnaire, and the profile offered no way to add or correct answers afterward.
They sign up, get five or six questions in, and then I never hear from them again.
- 01 setup
19 questions, one per screen13 were required. “Skip for now” ended all of them.
- 02 chat
Then a blank chatNo context, no suggested place to start.
- 03 answer
So the answers stay generalMy-Fi answers from a profile that is nearly empty.
- 04 profile
And nowhere to fix itRaw field names, and nothing can be pressed.
Ask enough to begin, then follow their goals.
I reviewed all 19 questions and kept four required ones. Province, income and age provided basic context; asking about goals early let us offer relevant follow-ups.
Does My-Fi need this for the first answer, or can it ask later?
We hadn’t tested whether these questions worked better in setup or in chat.
The founder’s financial advisor reviewed the final question list.
What brings you to My-Fi?
Pick everything that applies. We'll only ask about the things you choose.
Use their goal to suggest a first question.
I used the goals people had selected to suggest a first question. Choose budgeting, for example, and My-Fi offers to help build a budget.
Make the profile something people can correct.
Shorter setup depended on being able to add context later. I designed two ways: updates from the conversation and a profile people could edit themselves.
My-Fi could mistake a hypothetical for a fact. Someone asking “What if I moved to Alberta?” might still live in Ontario. I added a confirmation step before replacing a saved detail.
Save a detail My-Fi has never held. Ask before replacing one it has.
I replaced raw field names with a readable, editable profile. Edits reused the setup questions, keeping the work manageable for one part-time developer.
John is in Ontario, in the 26–35 range, earning $50–75k, saving toward a first home while paying down a student loan. More details will help My-Fi give more personal guidance.
- Rebuilt profile: edit a saved answer or see why an empty section matters.
More people asked a question, and profiles grew.
I defined measures before release; the developer pulled the counts. We followed 13 people on the old flow and 18 on the redesign for 14 days. The shorter setup and rebuilt profile shipped in May.
| Measure | Old flow | New flow |
|---|---|---|
| Reached the chat | 9 of 13 | 16 of 18 |
| Sent a financial question | 9 of 13 | 15 of 18 |
| Median time to that question | 6m 30s | 1m 45s |
| Sent a follow-up in the same session | 5 of 13 | 11 of 18 |
| Came back within 14 days | 4 of 13 | 9 of 18 |
| Profile grew after setup | 5 of 13 | 16 of 18 |
What I'd do differently
Deciding what to ask turned out to be the smaller half. People also need to see what’s been saved and fix it when it’s wrong. Next time, I’d design that alongside the first questions.
Have questions?
Who did you test with?
An audit of the existing flow, a review of how other financial products introduce questions, and the founder's account of where people stopped. Five returning participants walked through the redesign afterwards, and the founder's financial advisor reviewed the final question list. We didn't run sessions with people outside the founder's network, which is the thing I'd prioritise next.
What shipped, and what didn't?
The four-question setup, the optional follow-ups and the rebuilt profile shipped for Web Summit in May. The inline confirmation prompt shipped in a simpler form. The goal-driven suggestions on the home screen were designed and hadn't shipped when my engagement ended. My-Fi has continued to evolve since.
One part-time developer was also rebuilding the frontend from Flutter to React. Reusing setup components for profile editing helped keep the work manageable.