Overview Problem Research Design Impact

Case Study 04 · Fintech · Conversational UX

4 min read

WhatsApp Integration

Applying for a home loan is intimidating, especially if it is your first time. Plenty of people stop partway through. The bank’s only way back to those people was an SMS and an email that arrived days later, by which time the moment had passed.

WhatsApp conversation from Absa Home Loans — a re-engagement message congratulating the customer on starting their Home Loan journey, with a Yes, let's go reply button.
01 — Overview

Overview and Impact

My Role
Senior Product Designer
Timeline
3 Month Proof of Concept
Platform
WhatsApp + Digital Sales Platform (Web)
Status
Concluded

Problem

The bank’s re-engagement ran on SMS and email, and both only fired halfway through the application journey. Anyone who stalled before that point was never contacted at all. No reminder, no offer of help, nothing.

Outcome

0%Uplift in home loan application completion rate
0Meta-approved conversation templates
3Months, a Proof of Concept concluded on schedule
02 — Problem

The Challenge and the Objectives

The Challenge

Most home loan applicants don’t abandon because they’ve changed their mind. They abandon because the moment gets hard — a document they don’t have on hand, a field they’re not sure how to answer, a step that needs a detail they’ll have to go and find.

Existing re-engagement couldn’t reach those moments. It fired on a schedule, not on behaviour.

Objectives

  • Reach people at the moment of friction — not at a fixed point in the funnel.
  • Match the message to the customer — different drop-off behaviour needs different content and timing.
  • Stay compliant and non-intrusive — every message governed by Meta’s Business Messaging Policy.
  • Prove it inside three months — scoped as a Proof of Concept with a defined window.
03 — Research

Discovery and Research

What I did

Before designing anything, I mapped the existing journey end to end — every digital touchpoint, and where in the funnel the SMS and email triggers fired. Then I went through the home loan application analytics page by page, rather than looking at the funnel as a single conversion number.

The question wasn’t just when customers dropped off. It was where intervention would actually be effective. Usability testing ran alongside the analytics, so both drop-off points were backed by quantitative and qualitative evidence.

Key insights

Two pages surfaced immediately as the highest-friction moments in the entire application.

  • Property Details — applicants without an offer to purchase stalled here, with nothing prompting them to return once they had one.
  • Document Upload — documents had to be gathered offline, so the application stalled while people went to find them.
The full as-is journey, with drop-off identified page by page rather than as a single funnel figure.
04 — Design

Conversational Design & Compliance

Designing the conversation

I designed 12 conversational pieces, each mapped to a specific moment and a specific customer state rather than a single generic nudge.

Three of the twelve:

Reminder nudges — early drop-off Voice of Customer feedback — disengaged users Check-ins — completed journeys

Working inside Meta’s policy

WhatsApp is Meta’s platform: every template — its exact copy, format and permitted interaction pattern — is submitted and approved before it reaches a customer.

Certain conversation types can’t be initiated freely, template copy has hard limits, and patterns that feel natural to a designer get flagged. Getting all 12 approved was a design discipline in its own right.

Push Notification screen — Where the conversation starts: a push notification.

01 — Push NotificationWhere the conversation starts: a push notification.

Chats List screen — A verified business account in the chats list.

02 — Chats ListA verified business account in the chats list.

Opening Message screen — A greeting and one tappable reply.

03 — Opening MessageA greeting and one tappable reply.

Buyer Type screen — One question that decides the branch.

04 — Buyer TypeOne question that decides the branch.

Type Picker screen — A native list sheet, not free text.

05 — Type PickerA native list sheet, not free text.

First-Time Branch screen — The first-time buyer branch, with an exit.

06 — First-Time BranchThe first-time buyer branch, with an exit.

Reference Number screen — The reference number, sent as its own message.

07 — Reference NumberThe reference number, sent as its own message.

Back to the Application screen — The handoff back into the application.

08 — Back to the ApplicationThe handoff back into the application.

↔ Drag to move through the conversation · 8 screens

05 — Impact

Impact and Lessons

Results

The Proof of Concept concluded on schedule and delivered the uplift it set out to test.

The two drop-off points identified at the start proved to be the highest-leverage moments to intervene, which validated the page-by-page approach over a single funnel view.

Lessons

Know Your Customer pre-population was scoped out, not designed out. Pre-populating the application for existing customers takes real work off the person filling it in, and it is the first thing I would activate in a full rollout. The uplift was achieved without it.

Conversational UX isn’t simpler than interface UX. Take away the screen and the complexity doesn’t disappear. It moves into language, timing and flow logic.

Live prototype

Have a go yourself

A workable prototype, running right here — not a video and not a slideshow.

  1. Open the notification.
  2. Tap into the chat.
  3. Pick a buyer type.
  4. See the branch that customer would actually receive.
Best viewed on desktop

The interactive prototype is desktop-only — open this page on a larger screen to try it.