Listing cards

Listing cards play a pivotal role in every stage of a buyer’s search journey. On initial landing, they orient buyers to Etsy’s inventory and help them decide whether additional refinement is needed. Once buyers are generally satisfied with the results, listing cards surface essential details that support item evaluation and comparison.

My role

Led the strategic and incremental design of search listing cards for Etsy’s web experience.

Impact

Modernized search listing cards, created cross-platform consistency, partnered with ML to optimize and introduce new signals, and meaningfully improved key business metrics including CR and PcCR.

Buyer problems

Overwhelming volume of results, irrelevant listings, and challening to differenciate items.

Outcome statement

As a buyer on Etsy, I can easily explore and evaluate results in search, so that I can successfully collect possibilities.

Listing card signals

Signals give buyers meaningful information that supports decision-making. In close partnership with design systems, analytics, and research, we simplified listing card templates, aligned them across platforms, and introduced a more modern visual language.

Digital download

Introducing this signal lifted GMS by $12.1M, CR by +0.89%, and PcCR by +1.3%. It resolved the long-standing issue of buyers being unable to quickly identify item formats.

Seller name

Streamlining the seller name and making it clickable improved quality clicks by +0.89% and PcCR by +1.66%. Elevating this element reinforced the human behind each shop while creating a clearer path to shop pages.

Matieral swatches

Introducing material swatches enabled buyers to quickly understand metal options for jewelry listings. This feature produced a +0.25% lift in PcCR and +0.73% in CR.

Signal removal

As part of a cross-platform standardization initiative, we simplified listing cards by removing lower-value signals. Eliminating the “sale-ends-in” signal alone lifted GMS by +0.84%.

Leveraging ML

Designing for a two-sided marketplace with millions of listings requires accommodating highly variable, unstructured data. By partnering closely with ML teams, we transformed raw data into structured, consumable formats that could be seamlessly integrated into the user experience.

Signal ranking

Using a multi-arm bandit, we developed a model that dynamically selected which signal to show in a dedicated slot. While this experiment was ultimately ramped down, it generated valuable insights into the balance between consistency and dynamic content.

Color swatches

We partnered with ML to build an LLM model that inferred and standardized color options for listings. Adding color swatches to listing cards met buyer expectations and improved key metrics: CTR by +0.75%, quality clicks by +0.15%, and AOV by +0.98%.

Highlights & learnings

UI doesn’t exist in isolation: The hierarchy of listing card elements depends on what surrounds them; even small adjustments can have outsized impact.

Less is more: Signals are important, but too many create visual and cognitive overload.

Design around real inventory: In a two-sided marketplace, effective solutions must be built around the realities of our unique inventory and data structures.

← Etsy Filters

→ ADP Developer Logs

Contact

Email: claire.casaregola@gmail.com
LinkedIn: @clairecas