TinySurprise
E-Commerce Gifting
Started as a product intern building mobile apps and grew into an APM role over six years at the founding team. Built India's personalised gifting platform from zero. Along the way: a Walmart-inspired design pattern that killed conversion, a routing logic problem that looked like a logistics failure, and a redesign that unlocked 3x YoY growth.
from 55% baseline
post redesign
The original design followed a pattern inspired by Walmart's less-cluttered 2009 UI experiment to minimal product display, heavy on category navigation, slow to surface recommendations. Intent: reduce overwhelm. Result: users couldn't find what they wanted and didn't convert.
The first version also put the 3-question onboarding (occasion → recipient → location) as a gate before any browsing to asking for context before giving value. Users installed, reached the recommendation screen, and dropped.
What the data showed: Drop rates were concentrated on the recommendation screen, not at install. Users were completing onboarding but not converting on recommendations. The flow asked for too much too early. In an earlier experiment with a simplified homepage to similar to Walmart's reduced-SKU experiment to we also saw 30 orders in 3 months: strong press, weak conversion. The pattern was clear: simplicity without relevance doesn't convert.
v1 to original design
v1 to category flow
v1 to recommendation screen
The fix: Flip the information architecture. Home screen leads with curated gift recommendations. The 3-question flow becomes a personalisation filter you invoke when you want precision to not a mandatory gate. Visuals-first, price-prominent, occasion-tagged browsing.
v2 to redesigned home
v2 to gift browse
v2 to occasion filter
v2 to personalised results
Standard hypothesis: logistics partners weren't reliable. My hypothesis after mapping delivery failures by pin code and time-of-day: the routing logic was sending orders to the wrong partner for the wrong zone at the wrong time. Urban pin codes have wildly different lead times depending on traffic corridors and time-of-day to we were treating them as identical. The fix: time-slot-based routing system, lead time logic by zone. 24-hour fulfilment went from 55% to 90% without changing a single logistics partner.
The lesson: When a metric looks like an external dependency problem, map it before assuming. The routing logic was the problem. The logistics partners were doing exactly what the system told them to do.