Product Case Study · Myntra Wishlist

Shoppers save the products they want, then can't decide between them

How a unified decision workspace could lift Wishlist-to-order conversion, with no discounts, cashback or other incentives.

2,055

public conversations analysed with AI

20

survey responses

6

user interviews

Role

Product Management · Product Discovery · User Research · AI-Powered Product Development

Built with

  • Next.js 16
  • TypeScript
  • OpenAI API
  • Tailwind CSS 4
  • Recharts
  • TanStack Query
  • Zod

An independent case study and prototype. It is not a feature shipped at Myntra, no controlled experiment has been run, and no conversion results are claimed.

Business Problem

Saved is not the same as ready to buy

The problem

Users save several products that could meet the same need, but those products differ in fit, quality, material, styling, reviews and reusability. The information lives on separate product pages, so users keep reopening them and rebuilding the comparison in their heads.

Fragmented comparative evidence delays commitment to one Wishlist finalist, preventing high-intent users from progressing toward purchase.

The business opportunity

Improve Wishlist-to-Bag progression and 30-day Wishlist-to-Order conversion by helping users make confident purchase decisions, without monetary incentives.

Research

AI-assisted discovery, then primary research

Hypothesis

AI review discovery engine

  • Analysed 2,055 public conversations from Google Play, the App Store, Reddit and complaint forums.
  • AI classification of themes: purchase intent, decision uncertainty, product information, price sensitivity, postponement and comparison behaviour.
  • Aggregated evidence into prioritized opportunity hypotheses. Decision confidence and product-information uncertainty looked most promising.

Built to generate hypotheses, not to establish root causes or estimate how common a problem is across Myntra's users.

Research finding

Primary research

  • Surveyed 20 users.
  • Interviewed 6 participants from the chosen behavioural segment: people with a real purchase need whose Wishlist decisions were still unresolved.

The problem wasn't finding products. It was deciding between them.

Users could have enough information to shortlist several viable products and still not know which fit them best. Different products won on different attributes, leaving trade-offs unresolved and sending users back into comparison loops.

Root Cause

No unified way to compare viable alternatives

Users lack a single way to compare viable Wishlist alternatives against their own purchase priorities. The experience shows information product by product, but the decision needs several products side by side.

  1. 1

    Repeatedly reopening product detail pages.

  2. 2

    Reconstructing comparisons across fit, material, reviews, styling and more.

  3. 3

    Delayed commitment because no single product clearly wins.

The critical friction sits in Evaluate → Choose → Move to Bag.

Solution

Help Me Choose

A decision-support workspace inside Wishlist for resolving a shortlist of 2–4 viable products. Rather than more reminders or more raw information, it synthesizes trade-offs around the user's purchase context and priorities.

Shortlist

Pick the 2–4 products you are genuinely considering

Users select viable alternatives for the same purchase need, so the comparison stays small and relevant.

Select a step to see what happens there. Proposed

Prioritization

Three interventions considered

Option A

Richer product detail pages

What it does
Improve fit, sizing, review and photo information on each product page.
Fit with root cause
Partial. Builds confidence in individual products.
What remains unsolved
Users still compare across products manually.

Option B

Wishlist reminders

What it does
Nudge users to revisit products they saved.
Fit with root cause
Weak. Brings users back but does not help them decide.
What remains unsolved
The underlying decision problem is unchanged.

Option C · Advanced

Help Me Choose

What it does
Compare viable Wishlist alternatives in one workspace and synthesize trade-offs around the user's priorities.
Fit with root cause
Direct. Targets the research-supported root cause.
What remains unsolved
Impact and effort are not yet quantified; needs an experiment.

Decision: advance Help Me Choose because it most directly addresses the root cause. The call weighed root-cause fit, potential impact, research confidence and implementation effort. Impact and effort were not yet quantified, so this is directional, not a numerically validated ICE score.

Success Metrics

Conversion without hurting trust

Primary business metric

30-Day Wishlist-to-Order Conversion

  • Share of eligible users who buy at least one qualifying Wishlisted item within 30 days.

Target-cohort metrics

Is the intended segment moving?

  • Target-cohort 30-day Wishlist-to-Order conversion
  • Target Wishlist-to-Bag progression

Leading and diagnostic

Is the feature working as designed?

  • Compare-to-Bag conversion
  • Median time from comparison start to first Bag addition
  • Comparison abandonment rate

Guardrails

Is purchase quality and trust intact?

  • Returns and exchanges
  • Order cancellations
  • AI-generated claim quality
  • Feature adoption and friction during shortlist selection

Experimentation

How rollout would be decided

The proposed design compares the target cohort against an appropriate control group. Rollout depends on three conditions:

  1. Condition 1

    Conversion improves in the intended user segment.

  2. Condition 2

    Returns, exchanges and cancellations do not materially worsen.

  3. Condition 3

    AI comparisons stay evidence-grounded, with no unsupported product claims.

If the eligible segment is too small, broaden the entry point.
If users abandon the flow, simplify shortlist selection.

Scale only if conversion improves while purchase quality, trust and user experience stay stable.

Evidence Status

What is known, proposed and measured

Research finding
  • Survey of 20 users and 6 interviews with the target segment.
  • Users can shortlist several viable products yet struggle to choose between them.
Hypothesis
  • AI discovery themes: decision confidence and product-information uncertainty.
  • Cross-product comparison around stated priorities will speed commitment.
Not yet measured
  • No controlled experiment has been run.
  • No conversion lift, returns or cancellation results exist.

Implementation

Prototypes, not production

I built the AI discovery engine and a Help Me Choose MVP as prototypes to test the thinking. They are separate from any production system at Myntra and are covered in the slides.

  • Next.js 16
  • TypeScript
  • OpenAI API
  • Tailwind CSS 4
  • Recharts
  • TanStack Query
  • Zod

Reflection

Key learnings

  • A strong opportunity can exist after user interest is already established.

  • More information does not always solve a decision problem; users may need help synthesizing evidence they already have.

  • AI-assisted analysis can turn large volumes of qualitative feedback into testable hypotheses, but primary research is needed to validate the problem.

  • Prioritize by the mechanism behind user friction, not by how attractive a feature looks.

  • Measure business outcomes alongside trust, returns, cancellations and friction.

View the full slides →