Leaderboards
Savers were ranked against each other and their family on one board. Meant to motivate, it read as a race nobody had signed up for.
Product DesignUX ResearchUI Design
Savely
I A/B tested gamified saving against a plain control. Activation fell 7% and setup completion 6–8%, so I reverted it.
Savings setup completion under gamification — why I reverted
First-week activation with leaderboards + family nudges
Retention — no lift despite the added social pressure
Control vs Octalysis-driven gamified flow, read on setup completion + first-week activation
How might we …
Test whether shared progress changes how consistently people save, without exposing anyone's money to their peers?
Key screens
The two variants from the A/B test. Version B layered leaderboards, group quests and badges onto the same private savings app.
The experiment
Savely is a micro-savings app that sets money aside as people go about their day, without them having to think about it. I ran one experiment on it, as an A/B test: would gamification, meaning leaderboards and family-based nudges, get people to save more through social motivation?
Research
Saving feels quiet, private and pressure-free. The product never makes falling behind feel like failing.
Before designing anything, I ran a short questionnaire to understand how people save and manage money. It covered four areas: mobile usage, mobile-banking familiarity, money-management habits, and awareness of their children’s spending.
Key findings
People wanted a clearer picture of their cash flow: how income and spending change over time.
Saving was a priority, and they wanted more control over family-related expenses.
The second finding shaped the experiment. Family already influenced how people spent. Could it also influence how they saved?
Introduce some ease-in/ease-out functionalities after each component. Once a component is completed and they go to the next component in the scroll, introduce some ease-in/ease-out functionalities within the case study itself as each component comes into view.
The test
I built two versions. Version A was the standard app, as the control. Version B added gamified, incentive-driven elements, designed with the Octalysis framework: eight core drives that shape how people engage, used to decide which emotion each element should evoke.
What I tested
Leaderboards comparing savings progress
Family-based nudges encouraging collective saving
Messaging framed around shared goals and visibility
What I measured
Savings setup completion
First-week activation
Designing the experiment, not just the screens.
Testing a social-accountability idea cleanly meant trading signal strength for privacy and rigor. Each row is the call, what it cost, and why it held up.
DecisionWhat it costWhy it was worth it
01
Make social accountability the core bet
Sharing financial behavior raises real privacy concerns.
The test targets the one lever most likely to move savings behavior, not a safe tweak.
02
Clean A/B split on shared progress
Smaller samples, so significance comes slower.
Results are trustworthy enough to actually ship on — rigor over speed.
03
Group visibility stays opt-in
Weakens the social signal the test measures.
Consent is preserved, so the feature can survive contact with real users and reviewers.
04
Gentle nudges over binding pledges
Softer behavior change for far lower risk.
Users are not punished for slipping, so drop-off from guilt stays low.
05
Measure consistency, not totals
Undervalues large but irregular savers.
The metric rewards the habit the product is trying to build, not one-off deposits.
Savely: three social mechanics, one clean test.
Version B added three social mechanics to a private savings app. Each card shows what it looked like and what it did to the numbers.
Shared family quests
Family goals became group quests with each member's contribution on show. Support turned into accountability: one tester worried about letting their spouse down.
Badges on the home screen
Version B put badges, unlocked insights and quest counts above the same goals as Version A. Against the control, setup completion fell 6–8% and first-week activation fell 7%.
What ships now, and what has to earn its place.
One live surface was feasible today. Letting the model act, not just advise, waits on trust it hasn't earned yet.
- Now
Plan on the forecast
A live view per role, with AI insights you can simulate before acting.
Live in V1 - Next
Forecast earns its trust
Confidence and reasoning on every call; the wrong-prediction state designed first.
Feasible now - Later
Predictive auto-staffing
The model pre-fills shift changes, once it has earned the trust.
Gated on model trust
What happened
Trust in the model was the whole product, and I still designed the happy path first.
What I'd do differently
Prototype the wrong-prediction state on day one, not patch it in late.