Skip to main content
B2B SaaS Product discovery Data visualisation

Designing a data-driven dashboard for e-commerce returns

From an ignored screen to the product’s activation feature

Bounce rate on the dashboard −32%
Data requests to Customer Success −18%
Brands connected to exchanges +12%

When I joined, the dashboard had been built by developers without user research. It showed data like CO₂ savings and pending exchanges — metrics that looked impressive but had no impact on how brands actually managed their returns. The bounce rate was 80%. Brands were logging in, seeing nothing useful, and leaving.

Where
Spain
Company
Returns management startup
Year
2022–2024
My role
Product Designer — research, wireframing, UI design and documentation, working with the CEO, a Product Manager and frontend & backend developers.
Overview of the redesigned returns analytics dashboard
01

Context

The startup was in a critical phase of market adjustment in 2023, searching for product-market fit and needing to drive growth in both traction and revenue. The AARRR funnel analysis revealed a clear gap at the activation stage: users were reaching the product through sales demos, but they weren't discovering the value proposition. There was no "aha moment."

At the revenue level, the company needed brands to pay more. And at the operational level, over 200 brands were using the service, with 70% requiring returns data — but they couldn't access it themselves. Every report had to be manually requested through Customer Success, which meant developer intervention and constant delays.

AARRR funnel analysis showing the activation gap: users arrived through sales demos but never discovered the product's value
02

Before I Joined (2020–2021)

The original dashboard was a "Control Tower" built entirely on stakeholder feedback, not research. It displayed nine metric cards — reduced kilometers, saved CO₂, fully recovered units, backlog, daily saturation, validation delays, pending refunds, pending exchanges, and pending reviews.

The problem: none of these metrics helped brands make decisions about their returns. The data was based on what the company thought was interesting, not what users actually needed.

The original "Control Tower" dashboard: nine metric cards built on stakeholder feedback rather than research, which drove an 80% bounce rate
03

First Proof of Concept (2022)

When I started, I created a static no-code dashboard as a first proof of concept. It included products processed, revenue recovered, value from original price, and recycled product data alongside a stock grading table.

This was still based on ideas from different stakeholder groups rather than research — but it was enough to validate that brands wanted more from the dashboard.

2022 static no-code proof of concept: products processed, revenue recovered, value from original price and recycled product data, alongside a stock grading table
04

Discovery

In 2023, I led a proper product discovery process. First, I had to generate consensus within the team on the need for user research — the company had never done it before.

Research approach

Without direct users, I conducted stakeholder interviews with marketing, sales, and customer success teams to understand business metrics and impact. I ran expert interviews with the tech team to understand technical constraints before designing. Desk research covered competitor dashboards and data visualization best practices. Heuristic evaluations identified usability issues in the existing product.

With users, I ran 5 usability testing sessions with heatmaps and session recordings to observe real behavior. I conducted user interviews and card sorting exercises to understand mental models. A/B testing validated design decisions with data. Surveys captured quantitative feedback at scale.

Research methodology framework mapping methods across two axes: behaviours vs attitudes, and qualitative vs quantitative — with a separate column for the no-user context (stakeholder and expert interviews, desk research, heuristics)

Design process

We conducted interviews with key brands, competitor analysis, and data visualization research. We analyzed metrics to understand the current situation. We designed wireframes and conducted internal workshops to understand the vision and roadmap.

Design process: interviews with key brands, competitor analysis and data visualization research, metrics analysis, wireframes and internal workshops to align on vision and roadmap

Key findings

Brands need data on retained revenue and rapid detection of garment manufacturing faults. They want the ability to download data for generating detailed reports. They need to quickly and effectively acquire customer feedback. And they struggle when Customer Success provides reports in CSV format — they want visual, self-service data.

Pain points and user needs identified in research: retained revenue data, rapid detection of manufacturing faults, downloadable reports, faster customer feedback, and self-service visual data instead of CSV files from Customer Success

Problems identified

More than 200 users relied on the service, and 70% needed data on returns. But there was no automation — brands had to request data through Customer Success, causing manual work for the development team and delays. The company was losing potential customers during sales demos due to limited data accessibility. And brands still had more refunds than exchanges.

05

First Iteration

Based on the research, we decided to build two separate dashboards addressing different user needs. In collaboration with the development team, we agreed to implement only bar and line charts in the first MVP to ensure speed of delivery.

Client Dashboard

Focused on business outcomes: transport control, revenue retained, and customer feedback (NPS).

Client Dashboard full view: transports on the way and pending refunds at the top, revenue retained with a trend chart in the middle, and portal and refund-process NPS feedback with rating charts and customer comments at the bottom

Transport & Refunds Overview — a daily snapshot showing total transports on the way (split by drop-off vs. pickup with incident ratios) and total pending refunds (broken down by exchanges, store credit, and original payment). This gave brands an at-a-glance view of their most urgent operational items.

Transport and refunds overview: total transports on the way split by drop-off and pickup with their incident ratios, alongside total pending refunds broken down by exchanges, store credit and original payment

Revenue Retained — information on the total value retained by the platform (sum of exchanges and store credits). Brands can check current values and historical trends, with a graph showing the variation of returns by method over time. Time range filters let them analyze specific periods.

Revenue retained section: total value recovered through exchanges and store credit, with a line chart comparing retained revenue against returns over time and a date-range filter

Customer Feedback (NPS) — portal satisfaction scores using Net Promoter Score, with a rating distribution chart and individual customer comments. This gave brands direct visibility into how their customers experienced the return process — without waiting for a report from Customer Success.

Portal feedback section: Net Promoter Score with the detractors, neutrals and promoters split, a rating distribution chart and individual customer comments

Operations Dashboard

Focused on operational insights for internal teams and advanced brand users.

Operations Dashboard full view: refund processing times per stage, return and refund method distribution, top return reasons with customer comments, most returned products and returns by geography

Refund Times — average processing times for each stage of the return: 2 days from order confirmation to transportation, 3 days from pickup/drop-off to warehouse reception, 4 days from reception to refund. This helped brands identify bottlenecks in their return pipeline.

Average refund processing time per stage: 2 days from order confirmation to transportation, 3 days from pickup or drop-off to warehouse reception, and 4 days from reception to refund

About the Returns — operational data including total returns, total products, most used return methods, and refund method distribution. The top 5 return reasons and customer comments gave brands actionable insights to improve their products and sizing guides.

Operational returns data: total returns and products, most used return methods, refund method distribution, and the top five return reasons with customer comments

Most Returned Products — a chart showing which specific products had the highest return rates, allowing brands to identify manufacturing issues or misleading product photos.

Bar chart of the products with the highest return rates, used by brands to spot manufacturing faults or misleading product photos

Returns by Geography — breakdowns by country and region, relevant for brands selling internationally to optimize transportation automation and pricing configuration.

Returns broken down by country and region, used by brands selling internationally to optimise transport automation and pricing configuration
06

Second Iteration

After launch, I iterated based on user feedback and analytics: bug fixes from the first release, improved graph interactions with hover-over tooltips showing exact return counts, and flexible time range filtering with a date picker for custom periods.

Improved graph interaction: hovering a bar in the most returned products chart now shows a tooltip with the exact figure — VESTIDO LOLITA BEIGE, 9 units Flexible time range filtering: a date picker lets brands select a custom period — here Apr 22–29, 2024 — alongside preset intervals and an export option
07

Third Iteration (WIP)

By May 2024, brands needed more granular detail — specifically, the ability to see return reasons broken down by individual product, not just at the aggregate level.

08

Product Communication

The first MVP was communicated through targeted emails to active brands, featuring a "What's new?" announcement with a visual preview of the dashboard and a direct link. A second dispatch targeted brands in the closing process, using the dashboard as a sales tool.

"What's new?" email announcement sent to active brands, with a visual preview of the new dashboard and a direct link to it
09

Reflection

The dashboard went from an 80% bounce rate to becoming the product's activation feature — the thing that made brands understand the value of the platform. The 18% reduction in Customer Success requests meant the team could focus on closing new clients instead of pulling CSV reports. And the 32% bounce rate reduction proved that when you show users data they actually need, they come back.

What made this project work was insisting on research in a company that had never done it. The hardest part wasn't designing the dashboards — it was convincing the team that building features based on stakeholder opinions wasn't enough.

Contact me and tell me what you need

← Back to all projects