Online Marketplace

As a Senior Product Designer at a large online retailer, I noticed that after a cart redesign, conversion from Cart to Checkout had dropped by 22%, despite increased engagement. I used GA and ContentSquare to investigate and formed 3 hypotheses:

1. fulfilment options were visually disconnected from the checkout CTA
2. error messages were unclear and
3. mixed fulfilment options caused confusion.

I tested each change separately with multivariant and A/B testing, heatmaps, and user interviews. The winning variant restored conversion to baseline, and combined improvements led to a 3% increase in checkout conversion overall.

After rolling out the improvement to all users I then repeated the process to introduce each of the other 2 variants slowly into production based on their successful outcomes. This was to ensure that the variants worked together on the Cart page and not just in isolation.

Apple Watch App

Our pilots couldn't use the Logten iOS app due to cabin space restrictions, wifi signal inconsistency and lack of multitasking focus.

Pilots also didn't want to spend time retroactively entering flight times when they were off duty.

We had a backlog of features and solutions that were ready to be tested on the Apple Watch, so to expand our device support to include Apple Watch, I created a launch feature that solved a major friction point.

After studying the Apple Human Interface Guidelines (HIG), interviewing pilots and creating hypotheses, I designed, tested and iterated on an extremely well received solution in the form of an Apple Watch app that allows the pilots to tap once to record all 4 flight times without having to take their eyes and focus off their instruments.

This solution:

• increased our app engagement by 5% in the first 3 months
• increased flight input accuracy by 35% and
• user acquizition by 12% with a CAC of 1/4 of the LTV.

Mobility App

For the Mobility App, I created 6 user flow variants of the trip-planning interface for a complex Multivariant test.

The purpose was to validate (or invalidate) an assumption that users will purchase tickets faster and more often if they have 2 real-time information data points namely, 'next arrival' and 'icon of transit mode location on map'.

I rolled out the test by splitting the user traffic to 2% per challenger variant for a total of 12% of our users.

The measures of success were based on:

• Engagement with the next arrival affordance
• Engagement with map transit icon
• Ticket add to cart rate
• Time to checkout
• Average order value
• Revenue per session
• Quantity per order
• Session start to checkout

Testing showed that one variant reduced 'Session start to checkout' time by 24%, increased 'Revenue per session' by 38% and improved user confidence in the route choices. We rolled out this variant as the final design based on these insights.