Behavioral DesignRetentionResearchStrategy

Auto-Reload for a Wallet Users Kept Leaving

Accrue × Westside Market2025

Users claimed the discount, made one purchase, and disappeared. I studied 15 retention loops and designed auto-reload around what worked.

Jump to solution

Product

Westside Market wallet (Accrue embed SDK)

Team

1 Designer, 1 PM, Product leadership review

Timeline

Sep 2025 · One-week design sprint

Role

Research, Strategy & Design

THE CLIENT

Accrue × Westside Market

Accrue builds white-label co-branded wallet and loyalty platforms for enterprise retail brands. Westside Market (WSM) is a regional grocer that uses the Accrue wallet for loyalty and payments.

By September 2025, the WSM wallet had run for months. Users claimed the welcome discount, made one purchase, then vanished. The product team needed a way to bring them back without touching the reward economics. The fix had to come from design and behavior, not money. The team handed me a one-week sprint to design auto-reload, the retention tactic with the most leverage. To do it right, I started with the research.

PROJECT GOALS

  • 1Design auto-reload as the headline retention tactic
  • 2Pull users from first purchase to second without changing the reward economics
  • 3Surface the cost of inaction without crossing into dark patterns
  • 4Recommend a posture, not just options

THE PROCESS

01

Research

Days 1–3

Studied 15 consumer apps that retain users (Starbucks, Duolingo, Sephora, ClassPass, Nike SNKRS, and ten others) across 7 dimensions each. The repeating patterns gave me a formula to design against.

02

Design & Recommendation

Days 4–7

Designed three auto-reload directions with explicit tradeoffs and two companion patterns. Recommended the trustworthy angle as the primary surface, with selective urgency layered on top.

THE PROBLEM

Users treated the wallet like a one-time discount.

Onboarding worked. Users claimed the welcome reward, made a first purchase, then disappeared. Between purchases, they forgot the wallet existed. I needed to give them a reason to come back, and auto-reload was the tactic with the most leverage.

Activity sparkline: gap between sessions, employee vs regular user overlay

01

~15 days between actions

Actions in the wallet averaged about two weeks apart, and most of that activity came from employees beta-testing. Regular users opened the app to buy something, then closed it.

Funnel: signups → first purchase → second purchase drop-off

02

Discount hunt, not habit

The welcome reward got users in. Nothing kept them. After the first purchase, users treated the wallet like a coupon and closed it.

Constraint diagram: locked economics, open design surface

03

No changing the reward schema

Economics stayed locked. Whatever I designed had to work inside existing earn rates and rules. My only lever was attention.

Re-entry map: 5 moments after first purchase, with sentiment labels

Tap dots to explore

04

Five moments where users might come back

I mapped every re-entry point: balance increase, purchase complete, reward expiry (30/7/1 day), drops or campaigns, purchase cancelled. Each carried its own emotional state and needed its own treatment.

THE RESEARCH

15 apps. Three patterns that mattered most.

I studied 15 consumer apps that pull users back, dissected across 7 dimensions each. Three patterns stood out as directly applicable to auto-reload. The full grid is in the gallery.

01

Starbucks Rewards

Starbucks Rewards: stars expiry + bonus challenge screens

Stars expire after 6 months. Bonus challenges ('buy 3 drinks in 5 days for 75 stars') create urgency windows without crossing into pressure.

31M+ active members~50% of US sales from RewardsExpiry + bonus challenges
02

Duolingo

Duolingo: streak counter + 'streak freeze' notification

No tangible rewards. Pure streak psychology. The Duolingo streak is loss aversion stripped to one screen.

83M+ MAUsStreak as the entire mechanicIndustry-leading edtech retention
03

ClassPass

ClassPass: credit balance + expiry warning

Engineered sunk-cost retention. Users paid for credits, the credits expire monthly, and users book classes to avoid the waste.

Monthly credit expirySunk-cost reactivationDirect parallel to wallet balance
💡

One pattern showed up across the 15 apps. I distilled it into a four-part formula and used it to design auto-reload: [Trigger] + [Auto-action] + [Loss frame] + [Perk]. The formula outlived the sprint and became the team's retention reference.

AUTO-RELOAD: THE DESIGN

Three postures. Two supporting patterns. One recommendation.

I translated the formula into three auto-reload directions, each with explicit tradeoffs. Stakeholders picked a posture instead of arguing about copy. Two companion patterns ship alongside the auto-reload surface itself.

Before
Add Money
After

Trustworthy angle: auto-reload settings + first-load success screen

01

Safe & Trustworthy (Starbucks / Apple Cash model)

'When your balance drops below $10, we'll reload $25.' 'Never get stuck at checkout with an empty wallet.' Every reload includes a guaranteed prize. Users set it once. The behavior runs on its own from there.

Aggressive angle: countdown banner + 'lock in bonus' modal

02

Aggressive Loss Aversion (TEMU-inspired)

Flashing alert: 'Don't risk missing out, your balance is low.' Timer pressure: 'Set auto-reload in the next 24 hours to lock in your bonus prize.' The angle leans on urgency and FOMO. Converts faster short-term, risks feeling manipulative.

STEP 1 OF 3

Select payment method

Gamified angle: spin wheel + 'bonus spins for auto-loaders'

03

Gamified (Dunkin / Grab model)

Reload screen doubles as a spin-the-wheel. 'Skip reload = skip your chance to spin.' Auto-loaders get 2x spins each month. The angle adds variable reward, the slot-machine pattern in a softer skin.

Recommendation summary card: trust spectrum with chosen zone highlighted

04

The recommendation

I recommended shipping the trustworthy angle as the primary surface. Layer urgency tactics (expiry countdowns) on top where they inform a user. For a wallet that holds real money, the trust position beats the conversion play.

Before
Add Money
After

With/Without comparison card showing money state

05

Supporting pattern: 'With vs Without' card

Two states side by side at the decision moment. With Load: $25 → $26.25. Without Load: $25 → still $25, with a bold red 'Lost: $1.25'. Surfaces the cost of inaction in concrete dollars, without crossing into manipulation.

Dashboard module: cumulative missed-rewards counter

06

Supporting pattern: cumulative missed-rewards counter

Dashboard module: 'You've missed $18.40 in rewards this month.' Missing value over time hurts more than missing once. The counter updates weekly so the number moves.

THE IMPACT

3

Auto-reload directions

Safe & Trustworthy, Aggressive Loss Aversion, Gamified. Three postures with explicit tradeoffs so stakeholders could pick the brand position, not argue about copy.

15

Apps researched

Each dissected across 7 dimensions. The comparison grid became the team's retention reference.

Trustworthy

The recommendation

I recommended shipping the trust posture, with selective urgency tactics layered where they inform a user. The conversion math pointed elsewhere; the product's actual brand pointed here.

Auto-reload is in development on the trustworthy angle. The 15-app research and behavioral formula became our standing retention reference, applied to the work that came next.

LOOKING BACK

1

Research produces value through specifics.

'Use loss aversion' is not a design direction. 'Cumulative dollars missed, updated weekly on the dashboard' is a design direction. The sprint mattered because the output was specific.

2

I chose the trust position over the conversion math.

TEMU's aggressive angle converts faster short-term. Starbucks's trust model holds longer. For a wallet that holds real money, I picked the trust position. The recommendation matched the product's actual brand and the user's expectations of a financial tool.

3

I drew a line on the trust spectrum.

Loss aversion can become manipulation. The sprint needed an explicit line for when urgency informs users versus exploits them. I plotted the tactics on a spectrum so later decisions had a reference instead of a vibe.

WHAT I'D DO DIFFERENTLY

I'd A/B test the three angles instead of recommending one from research alone. Even a directional read in this audience would have replaced my inference with data.

LOOKING AHEAD

Auto-reload is in development on the trustworthy angle. Next sprints: instrument the re-entry points, ship the cumulative-missed-rewards counter, test the With/Without comparison card at checkout.