Tallyway · 2025
Halving time to first expense report for a spend tool
Rebuilt the first-run experience of an expense tool so new finance admins reach their first approved report in one sitting, not three.
- Client
- Tallyway
- Year
- 2025
- Role
- Lead product designer
- Duration
- 10 weeks
- Services
- Product research
- Onboarding design
- Prototyping
The problem
Tallyway sells expense management to companies with 50 to 500 employees. Trials were healthy at the top of the funnel, but only 22% of new workspaces submitted and approved a first expense report within 14 days, and that single event predicted conversion better than anything else in their data. The existing setup asked admins to configure policies, approval chains, accounting sync and card feeds before they could see a single expense move through the system.
I started with the evidence already on hand: 18 months of product analytics, 340 onboarding-related support tickets and the sales team's call notes. Then I ran 11 interviews with finance admins, split between workspaces that activated and ones that stalled. The pattern was blunt. Stalled admins reached the accounting sync step, needed credentials from someone else, and left the tab open for a week. Nobody was confused by the product; they were blocked by a dependency we had put first.
Process
- 01
Moved setup behind the first win
I reordered onboarding around one goal: get a real receipt approved in the first session. Policies and approval chains got sensible defaults the admin could review later, and accounting sync became a checklist item with an 'invite your accountant' path. The trade-off was real. Some workspaces would run for days on default policies, so we added a visible banner and a list of anything approved before the rules were confirmed.
- 02
Prototyped against real receipts
Instead of polished screens, I tested a clickable Figma prototype seeded with each participant's own anonymised receipts. Seven usability sessions surfaced two issues the flow diagram hid: admins wanted to try mobile capture themselves before inviting employees, and the word 'policy' worried smaller teams. We added a 'snap a receipt with your phone' step using a QR code and renamed the step to 'Spending rules'.
- 03
Shipped in three measured slices
Engineering capacity was two developers for six weeks, so we shipped in slices behind feature flags: reordered steps first, then defaults, then the mobile handoff. Each slice ran against a holdout of 20% of new trials. That slowed the full rollout, but it told us which change did the work, which mattered when the team later debated removing the QR step to save a screen.
Selected images
We had argued about onboarding for a year. Samuel Mensah put the stalled-admin interviews on one wall and the argument was over in an afternoon.
Outcome
Within two months of full rollout, first-report activation rose from 22% to 41%, and median time from signup to first approved report fell from 4.2 days to 38 minutes. Onboarding tickets dropped by a third, mostly because accounting sync questions disappeared from the first week. The defaults banner became the most-clicked element in week two, which is exactly when we wanted policy work to happen.
- 41%
- first-report activation (22%)
- 38 min
- median time to first report
- −34%
- onboarding support tickets
Keep reading
All work