Myntra: closing the price-confidence gap

The business metric: the share of wishlisted items bought within 30 days of being added. Why don't they convert?

Between adding to a wishlist and buying, an item can leak in three places. The user never comes back to it (attention leakage). They come back and decide no (intent resolution). Or they still want it but something blocks them (decision leakage).

I scoped out intent resolution: someone consciously deciding no is healthy self-filtering, not a failure to fix. Within decision leakage, some blockers are outside Myntra's control, like waiting for a paycheck. One isn't: not knowing whether the price will drop.

An AI engine I built scanned Play Store, App Store and YouTube reviews. It found a real fit-and-size signal that matched my interviews, but almost nothing about pre-purchase hesitation, because people review after buying or complaining, not while quietly waiting. So primary research had to carry the work.

Across 4 interviews and 44 survey responses, waiting for a price drop was the single largest reason cited: 16 of 35 survey answers mentioned price, and 2 interviewees checked prices across apps before buying.

How strong is this? Thin and skewed: 4 interviews (short of my 5 to 6 target), mostly aged 18 to 24. The same person also showed different blockers on different items, so I treated each purchase decision, not each person, as the unit.

Myntra offers a price-drop toggle, but no context to judge whether today's price is good. So the rational move is to keep waiting. The want is already resolved; price confidence is the only open question. Meanwhile third-party trackers already fill this gap on Myntra's own platform, and Nykaa already messages users on WhatsApp when prices drop.

A WhatsApp message the moment a wishlisted item genuinely drops in price, showing the old and new price. Tapping it opens the product page directly. Every verdict is computed from the product's real price history, never a prediction: more than 15% below average is a "Verified Low Price", around average is "Typical Price".

Real example: chinos at ₹2,600, their lowest recorded price, against an average of ₹5,062. The verdict reads "Verified Low Price, you're saving ₹2,462". The prototype covers 10 real products, with price data from a third-party tracker; in production it would run on Myntra's own first-party data, which competitors can't copy.

• A price-history card on the product pageSame evidence, but self-serve. It only helps if the user remembers to come back, and attention leakage is half the problem.
• A personal buying agent that auto-purchasesDirectionally right, but very high effort and untested, far beyond this project's real infrastructure.
• One-tap autonomous orderingNot one interview pointed to checkout friction as a blocker.

• People think in sale events, not averages. So it's scoped as price transparency, not price prediction.
• It only helps people who are waiting for the wrong reason. So results are reported by zone, not blended.
• The lift could just be purchases pulled forward. The randomised test isolates real lift.