Experimenting with the Redbubble product page experience
The problem
Redbubble is a complex marketplace, with millions of product and artwork combinations. The product display pages didn't properly help customers to distinguish between these often highly similar products, leading to overwhelm and confusion, decreasing conversion and average order value from our highest traffic and revenue-generating page.
Research focus
I synthesised session replays, Kano analysis, new interviews, and existing research, which made it clear that the appraisal experience was too overwhelming for customers. The product range was not going to become any simpler, so we focussed in on improving comparability and clarity between products.
I created a future-state vision prototype to align stakeholders on this idea, including new ways to display information and altering the overall information hierarchy of the page.
Experimentation
Once alignment had been gained, we broke down our ideas to run targeted experimentation. We focussed on testing key hypotheses for effectively reducing friction, such as:
- Surfacing trust and support information earlier will increase customer confidence to purchase
- Reducing cognitive load during product comparison will reduce overwhelm
- Improved guidance and suggestions are more valuable than pure visual changes
This allowed us to improve conversion metrics incrementally while still learning valuable information for later in the project.
Quick-switch experiment
A key issue for customers was the number of product variants, and being unable to easily understand the differences between them. This is particularly relevant for t-shirts, where there are 19 different varieties. We wanted to experiment with providing more guidance.
After several iterations, I decided the ideal solution should allow customers to switch between multiple styles exposed on the page, as they could more quickly and easily understand popular products. This is also more similar to other shopping sites, and is therefore more familiar to customers.
However, based on my prior experience with Redbubble's product architecture, I knew an in-line multi-style experience wouldn't be technically feasible without re-architecture. Products at Redbubble are stored as separate entities rather than variants, meaning new data has to be loaded every time a product is selected. Knowing this, I designed a second, more technically feasible option upfront so we could keep momentum without blocking the team on feasibility.
Experiment outcome
- 4% of all t-shirt sessions interacted with the feature
- 60% switched to a different t-shirt style
- +2.5% average order value for t-shirt customers, as customers switched to higher priced shirts
- $2.6 million estimated annual revenue uplift
The experiment was rolled out to 100% of users, and extended to wall art and hoodies, which had similar comprehension issues. Conversion and AOV also increased in these areas.
Learnings
This work reinforced the idea that experimentation can deliver meaningful gains and learnings for the business. However, appraisal experiences at scale are ultimately constrained by the underlying product architecture — both the technical architecture of the Redbubble site and the structure of products in the catalogue. These results helped me and my team advocate for replatforming the product page, elevating our conversations and my role from isolated optimisations to influencing foundational product and platform decisions.