Refund Redesign

This project focused on redesigning the refund flow within HelloPrint's internal customer service platform — a critical process with direct impact on financial accuracy, customer experience, and operational efficiency.

The Problem

The refund flow was highly error-prone and heavily dependent on manual validation. Agents could initiate refunds without meeting key conditions, leading to frequent errors and rework.

This resulted in:

• Execution errors: refunds created without valid complaints or required conditions
• Lack of system validation: no checks for basic requirements (e.g., order cancellation status)
• High dependency on Team Leads: manual review required for most cases, creating bottlenecks
• Irrelevant options: outdated choices increased confusion and decision errors

The system relied on people to catch errors — instead of preventing them.

The Objective

Redesign the refund flow to reduce operational errors and eliminate unnecessary dependency on manual validations, moving responsibility to the system instead of the Team Leads. The goal: build a guided process that prevented errors before they occurred, increased agent autonomy, and integrated business rules directly into the experience.

To understand the source of errors in the refund flow, I analysed real workflows and collaborated closely with Team Leads to identify recurring failure patterns.

1. Lack of System Validation

Agents could complete refund requests without meeting basic conditions — no complaint created, or order not yet canceled. The system had no enforcement mechanism at any step.

2. Manual Validation as a Safety Net

Team Leads had effectively become a control layer — manually verifying each request instead of the system enforcing rules. This was unsustainable at scale.

3. High Operational Overload

Most cases required Team Lead review — including low-risk, straightforward scenarios — creating significant bottlenecks and slowing down the entire process.

4. One Flow for Different Scenarios

The system treated all refund types (complaint, cancellation, upgrade/downgrade) identically, despite each having different rules, requirements, and risk profiles.

5. Cognitive Overload from Irrelevant Options

Outdated and unused options increased decision complexity and made errors more likely — not from agent carelessness, but from a system that provided no clear path.

From manual validation to system-guided flow

I redesigned the refund flow by moving responsibility from Team Leads to the system itself — integrating business rules, validations, and automation directly into the experience.

1. Smart Validations — Error Prevention at Source

System validations now prevent incorrect requests before they can be submitted:

• Complaint-based refunds require a valid existing complaint
• Cancellation refunds verify that the order is actually canceled
• Invalid actions are blocked or redirected with clear, contextual guidance

2. Contextual Flows by Refund Type

Instead of a single generic flow, the system now adapts based on the scenario — Complaint, Cancellation, or Upgrade/Downgrade. Each flow activates only its relevant fields, specific validations, and corresponding business logic. Agents are no longer navigating a one-size-fits-all form.

3. Reduced Dependency on Manual Approvals

I redefined approval logic based on risk level:

• Low-risk cases (under €500) can be executed directly by agents
• Upgrade/Downgrade flows no longer require approval at all
• Only genuinely high-risk cases are escalated to Team Leads

4. Automation of Repetitive Tasks

Auto-fill of refund amounts in full refunds, removal of unnecessary fields, and elimination of redundant inputs — reducing manual effort and error risk simultaneously.

5. Improved Navigation Between Flows

Direct access to the complaint form from the refund flow, enabling agents to complete missing prerequisites without losing their context or having to restart the process.

6. Traceability & Control

Clear statuses (pending, approved, completed), logging of the responsible agent, visibility of actions and key dates, and the ability to delete incorrect requests before processing. The process now has a full audit trail.

• Autonomy vs. financial risk: Increasing agent autonomy introduced risk in handling refunds. → Clear thresholds: low-risk cases handled independently; high-risk require Team Lead validation.
• Existing workflow dependency: The process was deeply integrated into daily operations, limiting disruptive changes. → Improved the flow without breaking existing mental models, ensuring smooth adoption.
• Control vs. flexibility: Strict validations can block certain valid edge-case scenarios. → Designed a specific path without restrictions for exceptional cases, allowing agents to proceed when genuinely needed without compromising the main flow.

The redesign moved from a reactive process dependent on manual validation to a guided, more autonomous, and reliable system.

• Reduction of operational errors: Built-in validations prevented incorrect or incomplete requests at source — the most impactful change in the entire process.
• Lower dependency on Team Leads: Manual validation was significantly reduced, freeing Team Leads to focus on genuinely high-risk cases.
• Increased agent autonomy: Agents can now complete most refund scenarios independently, particularly for low-risk cases.
• Improved process efficiency: Eliminating unnecessary reviews and rework significantly reduced overall handling time.
• Greater consistency in decisions: Business rules are embedded in the system, reducing variability and human error across the team.

The system no longer relies on manual validation to ensure correctness — it enforces it by design.