Revamping Seller Voucher Creation for the Largest E-Commerce Platform in Southeast Asia

Introduction

Shopee’s seller voucher is one of the most important tools for growing orders, especially in Southeast Asia where price sensitivity is high. Unlike platform-funded discounts, seller vouchers shift the promotional cost to the merchants. This creates a scalable, lower-burn strategy to drive conversion. But internal data showed a steady drop in seller voucher creation across all seller types, from long-tail to top-tier sellers. Without action, Shopee would need to spend more to maintain growth, and smaller sellers would miss out on an important tool to compete.

I led the end-to-end redesign of the voucher creation experience across six Southeast Asian markets. The project began by investigating a key problem: sellers were creating fewer vouchers over time. Funnel data from the Seller Center showed significant drop-offs during the setup process, particularly when sellers encountered technical fields like discount type, minimum spend, usage quantity, and budget configuration.

Simulating the Voucher Flow with Maze

To evaluate the current voucher creation experience, I recreated the full seller flow in Maze and distributed it to real sellers. This allowed us to simulate a realistic task without our team's involvement. Sellers were asked to complete the flow just as they would in the actual app. From the Maze dashboard, we captured key performance metrics and behavioral insights, including success rate, drop-offs, misclicks, heatmaps, and qualitative feedback.

The results revealed friction across the journey. Although the flow had 11 steps, only 73.9% of participants successfully completed the task. Participants rated the experience an average of 3 out of 5, reflecting a neutral to slightly difficult experience. Participant comments highlighted confusion about parameter settings, lack of clarity on voucher benefits, and difficulty noticing key values like estimated budget. Heatmaps further supported these insights, showing heavy activity around form fields and inconsistent interaction patterns.

Maze Result - Old Seller Voucher Creation Flow

Maze Result - Old Seller Voucher Creation Flow Heatmap

User Interview

We conducted moderated user interviews using the same Maze prototype, guiding participants through each step while prompting them to share their thoughts out loud. This allowed us to observe their mental models, identify hesitation points, and gather in-depth feedback on decision-making and overall clarity. We used contextual inquiry techniques to understand real behaviors and captured responses across five key areas: familiarity, creation habits, goal setting, parameter interpretation, and perceived value.

To support analysis, we quantified participant responses by assigning 1 to 5 scores across these categories, making it easier to identify patterns, compare insights, and inform design decisions.

User Interview

Insights

From the combined data gathered through Maze tests and user interviews, we synthesized key usability themes and behavioral patterns. These are summarized below:

Proposed Solution 1: Smart Voucher

To reduce friction in the voucher setup process, I introduced a new voucher type called Smart Voucher, designed to simplify decision-making through AI-generated recommendations. Instead of manually filling in all the parameters like discount amount, minimum spend, and usage limits, sellers are now presented with a few ready-made options tailored to their store data. Each option shows the projected buyer increase, making the benefit of voucher creation more tangible and incentivizing adoption. Based on interview feedback, we also addressed the issue of hidden estimated expenses by surfacing them more clearly and using them as the core logic behind each smart voucher. Sellers can still customize any of the AI suggestions before publishing, maintaining flexibility.

Proposed Solution 2: Swipeable Recommendations

Building on the Smart Voucher foundation, I explored a more playful interaction model inspired by dating apps. Sellers are shown a stack of voucher recommendations with projected buyer growth and clearer estimated expense values. They can swipe right to accept and publish a voucher immediately or swipe left to skip. This interaction simplifies exploration, speeds up decision-making, and adds delight to what was previously a tedious form-based task. It transforms a complex, technical flow into a lightweight, engaging experience that lowers the barrier to entry, especially for casual sellers.

User Feedbacks

I ran another round of Maze surveys and user interviews to validate both proposed solutions. Overall, users appreciated the concept of AI-generated vouchers for reducing friction. Option 1, which focuses on visibility and quick selection of voucher types, showed higher success rate and lower drop-off, with participants noting how much faster and easier it felt. Option 2, which introduced a swipe interaction, received positive comments on intuitiveness but had a slightly lower success rate and more users expressing confusion. Both prototypes confirmed the demand for simplified voucher creation, with Option 1 performing better in clarity and completion.

Option 1 User Feedback

Option 2 User Feedback

However, users still expressed the need for more control over estimated expenses and preferred to base their final choice of voucher on that parameter.

Final Result

Based on the feedback, the final UI allows users to input their estimated expense upfront. Our platform then leverages internal data to generate optimized voucher parameters that stay within the given budget. Users are presented with multiple voucher options, each tailored to match the expense input, while still having the flexibility to select the one that best suits their campaign goals. This balances automation with user control and directly addresses the pain points identified in the research.