Michael Mendez
design.mikemendez@gmail.com
AI Comparison Summary
Role: Lead Product Designer
Team: Cross-functional (PM, Engineering, Stakeholders)
Timeline: ~3 months
Success metrics vs. control
Add to Cart: +11.56%
PDP View Rate: +.12%
Conversion: +25.22%
Average order value: +25.48%
problem
Comparing products on LG.com was an overwhelming experience. The amount of spec data involved made it difficult for shoppers to find what they actually needed to make a decision. By digging into the data, it became clear that most shoppers were consistently looking for the same types of information when comparing products.
process
We weren't confident how well this would work, so we decided to A/B test the feature before committing. We limited our test to the most compared products as a baseline. Using FullStory data and user recordings, we identified what was actually driving conversion and pulled that into a short AI-generated comparison summary. We set category-level guardrails based on the data, then let AI do the heavy lifting of translating spec-heavy information into clear, consumer-friendly comparisons. After one month of testing the results were hard to ignore.
We included a "Quick Pick" to help shoppers identify which TV matched their watching style, whether that was gaming, cinematic, or high-end design. This was followed by the four most sought after categories for TVs: picture quality, gaming performance, audio experience, and design. Each category broke down how the compared TVs stacked up within those areas. We rounded it out with a shared features section highlighting what was available across all compared products.
Success metrics vs. control
Add to Cart: +11.56%
PDP View Rate: +.12%
Conversion: +25.22%
Average order value: +25.48%
What I learned
The biggest challenge we ran into was flawed spec data, which led to inaccurate AI comparisons. It also became clear that specs alone weren't enough. Incorporating key features alongside spec data was essential for surfacing consumer-friendly benefits in a way that actually helped shoppers make a decision.
