# Scaling referrals with better UX
Rakuten Rewards · 2020 · Lead Designer · Case study
Source: https://albancarmet.com/case-study/rakuten-referral
Redesigning the referral flow to support growth and engagement.
In one line: More referral shares on mobile
Team: PM, 1 engineer and me · Duration: 2 months · AI: None

## At a glance

- Problem: Many members visited the referral page but never shared their link — a missed, cheap way to grow the member base.
- What I did: A survey of 654 members and a benchmark well beyond our competitors, then a mobile and a desktop redesign, each tested in two rounds.
- Result: Mixed: more members shared on mobile, slightly fewer on desktop, and page-to-share conversion fell on both. The old page was restored a few weeks later.
- Limit: I left two days before launch, and phase 2 — the growth lever — never shipped.

## The company

Rakuten Rewards (formerly Ebates) is Rakuten’s cash-back program in the United States. Members shop at more than 3,500 partner brands through the site, the app or the browser extension, and get back part of the commission the brand pays Rakuten.
- Sector: E-commerce · cash-back
- Members: More than 17 million
- My time there: 2018 – 2020 · Lead Designer

## The product: Referral

Referral lets members invite friends with a personal link, and pays both of them when the friend starts shopping: a cheap way for the program to grow.
- Share a link: From the referral page, the member sends a personal link: by text, by email, or by copying it.
- A friend joins: The friend signs up with that link and spends $25.
- Both get a bonus: $25 for the member and $10 for the friend. Marketing can change the amount with its campaigns.

## Process

Each step fed the next. None of them used AI: this was 2020.
- Listen: A survey of 654 members who had seen the page without referring anyone. (without AI)
- Look elsewhere: A benchmark well beyond our competitors, then product data with my PM. (without AI)
- Prioritize: An effort-versus-impact model that split the work into two phases. (without AI)
- Design and test: A mobile and a desktop redesign, each tested in two rounds. (without AI)
- Ship: Phase 1 launched; I left Rakuten two days before. (without AI)

## The problem — Members opened the referral page, then left without sharing

A large share of the members who opened the page never shared their link. Two metrics to move: more members sharing, and more page visits turning into shares.

## The approach — Build the missing customer insight before designing

There was no customer insight when I started, so I built it. What came back was about trust more than incentives: some members took the program for a scam.

## The trade-offs — Three decisions, and what each one cost

The first one shaped how the whole project was judged: phase 1 rebuilt trust, and the growth lever waited for a phase 2 that never came.

## What shipped — A page that explains, shares and tracks

Phase 1 rebuilt the foundations: explain the program, add sharing methods, track each referral, answer questions on the page.

## Constraints

- Inside the Rakuten design system: Every component had to come from, or fit into, the existing design system.
- A bonus marketing can change alone: The referral amount changes with campaigns: the design had to let marketing update it without new visuals.
- Reuse to ship fast: The email invite component and the existing email templates had to be reused as much as possible, to keep development light.

## Signals and research

There was no customer insight when I started, so I built it: a survey of members who had seen the page without referring, a benchmark beyond our industry, product data with my PM, then usability tests on each platform.
- Survey · 654 responses — Trust before incentives: Members who didn’t refer didn’t trust the program or understand it — some saw it as a scam. They also wanted to know where their invited friends stood. (654: answers: members who didn’t refer didn’t trust the program or understand it.)
- Benchmark — Looking outside our industry: Competitors offered nothing new, so I looked at referral programs from every kind of business. Ideas like social proof and leaderboards came from there. (Elsewhere: Competitors offered nothing new: ideas like social proof and leaderboards came from other industries.)
- Product data — Two methods did most of the work: With my PM, we found that SMS and copying the link were by far the most used sharing methods. The design gives them the most weight. (2: sharing methods, text and copied link, did by far the most work.)
- Usability tests · 2 rounds per platform — What testers didn’t get: On mobile, most testers didn’t understand the QR code; on desktop, some never saw the sharing methods, below the fold. Round 2 fixed both. (2 rounds: per platform: most testers missed the QR code on mobile, some the sharing methods on desktop. Round 2 fixed both.)

## Decisions and trade-offs

- Foundations first rather than gamification right away. Rejected: Social proof and a leaderboard now — Ship the gamification layer found in the benchmark in the same release as the redesign. Why: An effort-versus-impact model split the work in two phases. The first had to rebuild trust: explain the program, add sharing methods, track referrals, answer questions. What it cost: The growth lever: phase 1 was never designed to lift conversion on its own. In hindsight: Phase 2 never launched. Judged on conversion alone, phase 1 looked like a failure, and the old page came back a few weeks after launch.
- Drop the QR code rather than add one more sharing method. Rejected: Six ways to share — Text and copy link, plus email, Messenger, a QR code unique to each member and a generic share button. Kept: Fewer methods, bonuses in view — QR code and share button removed; the space goes to the methods people used, and earned and pending bonuses move up to the main page. Why: In round 1, most testers didn’t understand the QR code. Removing it, along with the generic share button, gave more room to the two methods that did the work. What it cost: In-person sharing, which answered the survey’s “limited sharing options” complaint. In hindsight: Round 2 solved every issue raised in round 1, and the desktop redesign started from this version.
- Keep the mobile / desktop difference rather than unify. Decided with: My PM. Rejected: The same methods everywhere — Align sharing across platforms: SMS existed only on mobile, email invites and contact import only on desktop. Kept: One structure, two sets of methods — Both platforms share the same page structure; each keeps its own sharing methods. Why: Neither my PM nor I could find why the platforms differed, and unifying them risked delaying the project. What it cost: A consistent sharing experience across mobile and desktop, traded for speed. In hindsight: After launch, sharing rose on mobile and dipped slightly on desktop.

## What shipped

- Referral tracking, friend by friend (Each invited friend as a card, with their step in the journey.): problem: Referrers couldn’t see where their friends stood, so they couldn’t nudge them — and the bonus felt out of reach. Solution: Each card shows Invited, Joined or Spent $25+, with a reminder link while the friend is on their way. The pending tab adds up the bonuses still to come.
- Answers on the page (A referral FAQ, so members don’t leave to find out.): problem: The survey surfaced recurring questions, and every exit to look for an answer was one less chance of sharing. Solution: A FAQ dedicated to referral, in an accordion, right on the Refer a friend page.
- Emails that nudge, without spam (The invitation, its reminder and the shop reminder, each sent from the friend’s card.): problem: Survey respondents cared about not spamming their friends, who already got plenty of emails from other brands. Solution: A new reminder for friends who joined but haven’t shopped yet, and a rule: no reminder within 7 days of the last one.

## Results

Mixed results, and a rollback a few weeks after launch.
(Feedback gathered after launch · I left Rakuten two days before it)
- Up: Members sharing their referral link, on mobile
- Down: Conversion from page visit to share, on mobile and desktop
- Slightly down: Members sharing their link, on desktop
- Rolled back: The old page was restored a few weeks after launch
Phase 1 was scoped to rebuild foundations and trust, not to lift conversion right away. That intent was never shared widely enough: when results came in mixed, nobody had a common frame to read them.
What worked:
- On both platforms, round 2 solved what round 1 had surfaced.
What didn't work:
- Page-to-share conversion fell, and the redesign was rolled back.
- I over-indexed on test findings: the referral activity section ended up more complex than it needed to be.

## What I took from it

Phase 1 needed its own definition of success. I should have pushed to agree on what success meant for phase 1, and written it down where everyone could see it. And I now use user feedback as a signal, not as a spec.