
Lead with the reward
The original picture-based asset explained the 30K reward for the referrer and the 15K discount for their friend, with a reassurance that the referrer would not be a guarantor.
CASE STUDY 02 · GROWTH · FINTECH
A growth-design story about making the value exchange clearer, removing hesitation and turning a referral feature into an experience people could confidently recommend.

THE STARTING POINT · 01
When someone invites a friend to try a financial product, they're putting a little of their own reputation on the line. They need to understand the offer, trust that the process is straightforward and feel comfortable explaining it to someone they know.
That was the opportunity behind Share & Earn at Numida. We wanted to make the referral experience easier to understand and more compelling to act on, without making it feel like a hard sell.
THE DESIGN QUESTION · 02
It was tempting to think of referrals as a button-placement problem. But making the action more visible wouldn't necessarily answer the questions in a customer's mind: Why should I share this? What's in it for my friend? What happens after I send the invitation?
So I approached the experience as a complete value exchange. The design needed to explain the benefit, make the invitation easy to find and set expectations for what would happen next.
THE ORIGINAL EXPERIENCE · 03
The original Share & Earn screen introduced the reward and gave customers a way to share their link. It explained the offer, but the experience left us with a bigger question: what would help someone feel confident recommending Numida to a friend?

The starting point for the redesign. This screen is shown separately from the later A/B test variants.
RESEARCH & EXPLORATION · 04
In a team brainstorm, we mapped customer concerns and possible ways to make the referral proposition more compelling. The discussion covered fear of becoming a guarantor, the effort involved in meeting loan requirements, and the value of showing that other people were already benefiting from referrals.
We grouped the ideas around customer fears, social proof, referral summaries and ways to make the value exchange easier to understand.
We translated those themes into lightweight wireframes and explored how the referral journey could explain the reward, invite friends, show sharing options and communicate what happens next. These explorations helped shape the final concepts we took into the A/B test.





EXPERIMENT DESIGN · 05
We used GrowthBook to compare three referral-page assets: the original reward-led control, a treatment that addressed customers’ fear of being responsible for a friend’s loan, and a treatment that used social proof to show that other customers were already referring friends and earning rewards.
The experiment began on 26 May 2022, with results analysed on 6 July. GrowthBook automatically assigned eligible customers to a variation. Customers had to be using app version 2.29, and anyone who viewed more than one variation was excluded from the analysis.

The original picture-based asset explained the 30K reward for the referrer and the 15K discount for their friend, with a reassurance that the referrer would not be a guarantor.

A vertical stepper addressed a key concern directly: customers would not be responsible for their friend’s loan.

This version led with community proof — 9,134 people had shared Numida last month — before explaining the share and earn steps.
EXPERIMENT WINNERMEASURING THE RESULT · 06
We measured whether people who saw a variation went on to tap Share and whether their referred friends signed up. The report compared each treatment against the original control.
Increase for social proof compared with control.
Increase for social proof compared with control.
Percentage increases are relative to the control variation, as reported in the July 2022 experiment analysis. They are not percentage-point changes.
The report recorded a 0% chance of share-button conversion being worse if social proof was selected, and a 0.42% chance of referral sign-up rate being worse. For comparison, the client-fears treatment had risks of 0.03% and 2.74% respectively; the control had risks of 12.33% and 12.94%.
THE LEARNING · 07
The winning variation did more than explain the incentive. It showed customers that other people were already sharing Numida and benefiting from it. In this experiment, that message performed better than both the original reward-led asset and the treatment focused on guarantor fears.
The July 2022 report recommended improving the social-proof vertical stepper and making it the primary asset on the referral page. The experiment gives us evidence for choosing that direction, while leaving room to keep improving the content and experience.
A good referral experience doesn't just ask customers to spread the word. It gives them a reason to believe the recommendation is worth making.